development of eco-driving control system for connected

Development of Eco-Driving Control System for Connected and

Automated Hybrid Electric Vehicles

by

Siyang Wang

A thesis submitted to the

School of Graduate and Postdoctoral Studies in partial

fulfillment of the requirements for the degree of

Master of Applied Science in Automotive Engineering

Department of Automotive, Mechanical and Manufacturing Engineering

School of Graduate and Postdoctoral Studies

The Faculty of Engineering and Applied Science

University of Ontario Institute of Technology (Ontario Tech University)

Oshawa, Ontario, Canada

April 2020

© Siyang Wang, 2020

ii

THESIS EXAMINATION INFORMATION

Submitted by: Siyang Wang

Master of Applied Science in Automotive Engineering

Thesis title: Development of Eco-Driving Control System for Connected and

Automated Hybrid Electric Vehicles

An oral defense of this thesis took place on April 9, 2020 in front of the following

examining committee:

Examining Committee:

Chair of Examining Committee

Dr. Amirkianoosh Kiani

Research Supervisor

Dr. Xianke Lin

Examining Committee Member

Dr. Haoxiang Lang

External Examiner

Dr. Jing Ren

The above committee determined that the thesis is acceptable in form and content and that

a satisfactory knowledge of the field covered by the thesis was demonstrated by the

candidate during an oral examination. A signed copy of the Certificate of Approval is

available from the School of Graduate and Postdoctoral Studies.

iii

ABSTRACT

Hybrid electric vehicles (HEVs) were designed as a potential solution to the ever-

increasing global problems of the energy crisis and global warming through flexibly

utilizing both fuel and electrical energy. Besides, the emerging technologies of connected

and automated vehicles (CAVs) have provided huge possibilities to push the boundaries of

HEVs even further and thus have been extensively studied. In this study, a bi-level MPC-

based eco-driving strategy for CAHEVs is proposed and designed to improve fuel economy,

reduce exhaust emissions while ensuring driving safety under the most common driving

scenarios. First, the HEV powertrain is modelled, and the real-time data sources are in the

intelligent transportation system (ITS) are introduced. Next, the multi-objective problem is

formulated with three goals, namely, driving safety, fuel economy and emission reduction.

The simulation is carried out on a map with realistic driving conditions. The results

demonstrate the effectiveness and robustness of the proposed eco-driving strategy for

CAHEVs.

Keywords: Eco-driving, energy management strategy (EMS), connected and

automated hybrid vehicle (CAV), hybrid electric vehicle (HEV), intelligent

transportation system (ITS)

iv

AUTHOR’S DECLARATION

I hereby declare that this thesis consists of original work of which I have authored.

This is a true copy of the thesis, including any required final revisions, as accepted by my

examiners.

I authorize the University of Ontario Institute of Technology to lend this thesis to other

institutions or individuals for the purpose of scholarly research. I further authorize

University of Ontario Institute of Technology to reproduce this thesis by photocopying or

by other means, in total or in part, at the request of other institutions or individuals for the

purpose of scholarly research. I understand that my thesis will be made electronically

available to the public.

Siyang Wang

v

STATEMENT OF CONTRIBUTIONS

Part of this thesis has been submitted for publication as:

S. Wang and X. Lin, “Eco-driving control of connected and automated hybrid vehicles in

mixed driving scenarios”, Applied Energy, 2020. (Submitted)

The writing of the manuscript, vehicle modeling, platform setup, test simulation conducted

in this work and the mentioned submission are completed by the author. Co-author

reviewed and provided technical support when required.

vi

ACKNOWLEDGEMENTS

I would like to thank my supervisor, Dr. Xianke Lin, for his consistent support and

skillful guidance throughout this challenging thesis.

vii

TABLE OF CONTENTS

ABSTRACT ............................................................................................................ iii

AUTHOR’S DECLARATION ..................................................................................... iv

STATEMENT OF CONTRIBUTIONS .......................................................................... v

ACKNOWLEDGEMENTS ........................................................................................ vi

TABLE OF CONTENTS ........................................................................................... vii

LIST OF TABLES...................................................................................................... x

LIST OF FIGURES .................................................................................................. xii

LIST OF ABBREVIATIONS AND SYMBOLS ............................................................. xvi

Chapter 1. Introduction ..................................................................................... 1

1.1 Background and motivation............................................................................ 1

1.2 Objectives and assumptions ........................................................................... 6

1.3 Contributions ................................................................................................... 8

1.4 Outline ............................................................................................................. 9

Chapter 2. Literature Review ........................................................................... 11

2.1 Introduction................................................................................................... 11

2.2 Hybrid electric vehicles .................................................................................. 11

2.2.1 Series Hybrids........................................................................................ 12

2.2.2 Parallel Hybrids ..................................................................................... 13

2.2.3 Series-Parallel Hybrids .......................................................................... 15

2.2.4 Driving Cycle .......................................................................................... 16

2.3 Connected and automated vehicles (CAVs) .................................................. 17

viii

2.3.1 Eco-Driving scenarios ............................................................................ 18

2.4 Literature review on control strategies for CAHEVs...................................... 20

2.4.1 Energy management strategies (EMSs) for HEVs ................................. 20

2.4.2 Control strategies for CAHEVs .............................................................. 31

2.4.3 Existing research gaps in eco-driving for CAHEVs ................................. 33

2.5 Summary ....................................................................................................... 34

Chapter 3. Model Development ....................................................................... 35

3.1 Modelling of Hybrid Electric Vehicles (HEVs) ................................................ 35

3.1.1 HEV Configuration ................................................................................. 35

3.2 Intelligent transportation system (ITS) ......................................................... 43

3.2.1 Vehicle-to-Vehicle (V2V) ....................................................................... 43

3.2.2 Vehicle-to-Infrastructure (V2I).............................................................. 44

3.3 Summary ....................................................................................................... 45

Chapter 4. Methodology .................................................................................. 46

4.1 Problem Formulation .................................................................................... 46

4.2 Model predictive control-based eco-driving control strategy....................... 47

4.2.1 Optimization objectives ........................................................................ 48

4.2.2 Model predictive control (MPC) ........................................................... 57

4.3 Rule-based EACS............................................................................................ 61

4.4 Summary ....................................................................................................... 67

Chapter 5. Results and Discussion .................................................................... 69

5.1 Experiment setup .......................................................................................... 69

5.1.1 ADVISOR ................................................................................................ 69

ix

5.1.2 SUMO .................................................................................................... 70

5.1.3 Traffic simulation map .......................................................................... 73

5.2 Simulation results on a highway-urban trip.................................................. 77

5.2.1 Trip Setup .............................................................................................. 77

5.2.2 Driving Safety ........................................................................................ 80

5.2.3 Energy consumption and emission reduction ...................................... 88

5.3 Simulation results on a suburban-urban trip ................................................ 90

5.3.1 Trip Setup .............................................................................................. 90

5.3.2 Driving Safety ........................................................................................ 93

5.3.3 Energy consumption and emission reduction .................................... 100

5.4 Summary ..................................................................................................... 101

Chapter 6. Conclusion and Future Works ....................................................... 103

6.1 Conclusion ................................................................................................... 103

6.2 Future Works ............................................................................................... 104

Reference ......................................................................................................... 106

x

LIST OF TABLES

CHAPTER 2

Table. 2. 1 HEV working modes ................................................................................ 22

Table. 2. 2 Summary on rule-based energy management strategies for HEV ............ 26

Table. 2. 3 Summary on optimization-based energy management strategies for HEV

..................................................................................................................................... 30

CHAPTER 3

Table. 3. 1 The studied HEV specifications ............................................................... 37

Table. 3. 2 V2V data for this study ............................................................................. 43

Table. 3. 3 V2I data used for this study ...................................................................... 44

Table. 3. 4 Listed traffic signal phases ....................................................................... 45

CHAPTER 4

Table. 4. 1 Parameters concerning the driving safety ................................................. 49

Table. 4. 2 Parameters for IDM .................................................................................. 62

Table. 4. 3 Parameters for rule-based EACS .............................................................. 65

CHAPTER 5

Table. 5. 1 Background vehicle specifications ........................................................... 75

Table. 5. 2 Specifications of the simulation map ........................................................ 75

Table. 5. 3 Parameters for simulation ......................................................................... 76

xi

Table. 5. 4 Specifications of the highway-urban trip .................................................. 78

Table. 5. 5 Results from the proposed MPC-based eco-driving strategy and the EACS

rule-based strategy (SOC corrected) on a highway-urban trip ................................... 90

Table. 5. 6 Specifications of the suburban-urban trip ................................................. 91


rule-based strategy (SOC corrected) on a suburban trip ........................................... 101

xii

LIST OF FIGURES

CHAPTER 1

Fig. 1. 1 Energy consumption by the end-user sector [5] ............................................. 2

Fig. 1. 2 Carbon dioxide emissions by the end-user sector [5] ..................................... 2

Fig. 1. 3 V2V and V2I communications for CAVs in ITS ........................................... 6

CHAPTER 2

Fig. 2. 1 The series hybrid powertrain ........................................................................ 13

Fig. 2. 2 The parallel hybrid powertrain ..................................................................... 14

Fig. 2. 3 The series-parallel hybrid powertrain ........................................................... 16

Fig. 2. 4 The Urban Dynamometer Driving Schedule (UDDS) ................................. 17

Fig. 2. 5 The classifications of control strategies for HEV ......................................... 21

Fig. 2. 6 The vehicle working modes, (a) electric only, (b) hybrid / electric assist, (c)

battery charging, (d) regenerative braking, (e) engine only ........................................ 23

CHAPTER 3

Fig. 3. 1 Single-shaft pre-transmission HEV Powertrain ........................................... 36

Fig. 3. 2 Operation points of Saturn 1.9L (63kW) SOHC SI Engine on fuel economy

and engine-out emissions contour maps (a) Fuel economy (b) HC emission (c) CO

emissions and (d) NOx emission ................................................................................. 39

Fig. 3. 3 Equivalent circuit model for battery pack .................................................... 40

Fig. 3. 4 Longitudinal vehicle dynamics ..................................................................... 42

xiii

CHAPTER 4

Fig. 4. 1 Illustration for driving safety parameters ..................................................... 50

Fig. 4. 2 Driving scenario classifier (DSC) workflow ................................................ 51

Fig. 4. 3 Sample cases at the TSCI without leading vehicles (a) brakes are applied due

to the red signal, (b) brakes are applied because the green signal phase duration is not

sufficient for passing (c) vehicle decides to drive across the intersection in 25s ....... 55

Fig. 4. 4 The framework of the proposed MPC-based eco-driving strategy ............... 59

Fig. 4. 5 Pseudo-code of the proposed algorithm ....................................................... 60

Fig. 4. 6 EACS rule-based illustrations ...................................................................... 64

Fig. 4. 7 Pseudo code of rule-based EACS ................................................................. 66

Fig. 4. 8 Workflow of the EACS ................................................................................ 67

CHAPTER 5

Fig. 5. 1 Sample ADVISOR interface for customizing vehicle parameters, a vehicle

with a parallel hybrid powertrain is configured .......................................................... 70

Fig. 5. 2 Sample SUMO-GUI interface, zoomed-out view of the Cologne, Germany

urban transportation network ...................................................................................... 72

Fig. 5. 3 A zoomed-in view of the Cologne, Germany urban transportation network [77]

..................................................................................................................................... 72

Fig. 5. 4 Urban network map of Cologne, Germany [77] ........................................... 74

Fig. 5. 5 The simulated highway-urban trip from point A to point B that passes point P

[77] .............................................................................................................................. 78

Fig. 5. 6 Speed profile by MPC-based eco-driving strategy in on an highway-urban trip

..................................................................................................................................... 79

Fig. 5. 7 Speed profile by the rule-based EACS strategy on an highway-urban trip .. 79

Fig. 5. 8 Speed profile by MPC-based eco-driving strategy with road speed limit

throughout an highway-urban trip .............................................................................. 83

xiv

Fig. 5. 9 Signal anticipation by MPC-based eco-driving strategy on an highway-urban

trip ............................................................................................................................... 83

Fig. 5. 10 Car-following scenarios by MPC-based eco-driving strategy on an highway-

urban trip ..................................................................................................................... 84

Fig. 5. 11 The MPC-based eco-driving strategy results including (a) Brake-specific fuel

consumption (BSFC), (b) Efficiency of the electric motor (EM) and (c) SOC of the

battery pack on an highway-urban trip ....................................................................... 85

Fig. 5. 12 The EACS rule-based strategy results including (a) Brake-specific fuel


battery pack on an highway-urban trip ....................................................................... 86

Fig. 5. 13 The distribution of engine working points by the MPC-based eco-driving

strategy and the EACS rule-based strategy on an highway-urban trip ....................... 87

Fig. 5. 14 The exhaust emissions of HC, CO, NOx produced by the MPC-based eco-

driving strategy and the EACS rule-based strategy on an highway-urban trip........... 87

Fig. 5. 15 The simulated suburban-urban trip from point A to point B [77] .............. 91

Fig. 5. 16 Speed profile by MPC-based eco-driving strategy on a suburban-urban trip

..................................................................................................................................... 92

Fig. 5. 17 Speed profile by the rule-based EACS strategy on a suburban-urban trip . 92


throughout a suburban-urban trip................................................................................ 95

Fig. 5. 19 Signal anticipation by MPC-based eco-driving strategy on a suburban-urban

trip ............................................................................................................................... 95

Fig. 5. 20 Car-following scenarios by MPC-based eco-driving strategy on a suburban-

urban trip ..................................................................................................................... 96

Fig. 5. 21 The MPC-based eco-driving strategy results including (a) Brake-specific fuel


battery pack on a suburban-urban trip ........................................................................ 97



battery pack on a suburban-urban trip ........................................................................ 98


strategy and the EACS rule-based strategy on a suburban-urban trip ........................ 99

xv

Fig. 5. 24 The exhaust emissions of HC, CO, and NOx produced by the MPC-based

eco-driving strategy and the EACS rule-based strategy on a suburban-urban trip ..... 99

xvi

LIST OF ABBREVIATIONS AND SYMBOLS

BSFC Brake-Specific Fuel Consumption

CAV Connected and Automated Vehicle

CAHEV Connected and Automated Hybrid Electric Vehicle

CF Car Following

CO Carbon Monoxides

EACS Electric-Assist Control Strategy

EM Electric Motor

EMS Energy Management Strategy

FD Free Driving

HC Hydrocarbons

HEV Hybrid Electric Vehicle

CE Internal Combustion Engine

ITS Intelligent Transportation System

MPC Model Predictive Control

𝐍𝐎𝐱 Nitrogen Oxides

SA Signal Anticipation

TSCI Traffic Signal Controlled Intersection

V2I Vehicle-to-Infrastructure

V2V Vehicle-to-Vehicle

1

Chapter 1. Introduction

1.1 Background and motivation

Since the birth of the first vehicle in the late 1880s, the automobile industry has been

rapidly expanding globally, and yet automobiles are playing a crucial and indispensable

role in modern civilization. According to the statistical report, more than 1 billion vehicles

are running on the ground on a daily basis. The development of automobile technology has

brought great convenience and economic benefits to human societies. Despite that,

however, some serious problems also have come into existence and become urgent global

concerns, such as air pollution caused by Greenhouse Gas (GHG) and depletion of

nonrenewable energy sources, especially in fossil fuel [1–4].

According to the 2020 annual statistical report by U.S. Energy Information

Administration (EIA), the total global energy consumption classified by end-use sector is

illustrated in Fig. 1. 1, and the total global energy-related carbon dioxide emissions

classified by end-use sector is illustrated in Fig. 1. 2 [5]. It is clearly shown that the

transportation sector, which consumed approximately 28 Quadrillion Btu (8.21 ×

1012 kWh), accounted for around 27% of total energy consumption in 2019 and is expected

to keep a steady amount of use through 2050. On the other hand, the carbon dioxide

emissions of approximately 1.9 Gt produced by the transportation sector accounted for

around 37% of total carbon dioxide emissions in 2019 and is expected to stay as the top

carbon dioxide producer through 2050. Meanwhile, the global trends of urbanization are

significantly increasing the demand for more automobiles in big cities, which will

inevitably deteriorate the problems of the energy crisis and exhaust emissions.

2

Fig. 1. 1 Energy consumption by the end-user sector [5]

Fig. 1. 2 Carbon dioxide emissions by the end-user sector [5]

3

In response to those problems, many researchers in the past decades have proposed a

variety of approaches, among which hybrid electric vehicles (HEVs) have shown

tremendous potential to save energy and reduce pollutant emission [6, 7]. HEVs, as

believed to be the transitional vehicles for automotive industry to migrate from

conventional internal combustion engine vehicles to pure electric vehicles, stand out to

address the problem by introducing a hybrid powertrain consisting of not only the

conventional internal combustion engine but also the electric motor/generator supplied by

the onboard battery pack [2, 8]. In addition, the regenerative braking technology utilized

by HEVs enables the recycling of waste kinetic energy on the wheels. Such configuration

design offers great flexibility in energy split control while ensuring the power demand from

the wheels [2].

Since the commercialized production of HEV by Toyota and Honda, there is still an

increasing effort to push for even better fuel economy and emission reduction from HEVs.

This can be achieved through using a well-designed powertrain, light material to reduce

aerodynamic resistance, optimal component sizing, and so on. Among those approaches,

the most effective and cost-friendly one is through optimal control of power flows between

the internal combustion engine and electric motor such that the powertrain can work in the

most efficient state while meeting the driving demand. Therefore, adopt the right energy

management strategies (EMSs) is crucial to the improvement of fuel economy and

reduction in exhaust emissions [2].

For several years, great effort has been devoted to the study of energy management

strategies (EMSs) for HEVs [3, 9, 10]. They can be divided into two main categories,

optimization-based strategies and heuristic/rule-based strategies. In heuristic/rule-based

strategies, a set of rules designed by human expertise, intuition and/or mathematical models

are deployed in real-time to determine the control action for each time step [11]. Without

taking into account the distinction in driving conditions, they can end up with results that

have poor adaptability and accuracy. The optimization strategies, on the other hand,

introduce numerical or analytical control strategies to derive the optimal actions [2]. They

can be further categorized into global optimization methods and instantaneous optimization

4

methods. As a commonly used method of global optimization, dynamic programming (DP)

can ensure the global optimality by being able to access the full information of the driving

conditions a priori [12, 13]. Thus, it is often used as a benchmark method for many other

EMSs. In real-world trips, however, it is impossible to obtain complete knowledge in

advance. In addition, the complexity of DP usually leads to high computational cost, which

is not suitable for the online control scenario. Pontryagin’s minimum principle (PMP), as

another popular global optimization method for HEVs, as compared to DP and the results

showed that they were very close to the ones calculated by DP [14]. However, PMP is

optimized concerning a certain driving cycle and hence cannot guarantee the optimality of

other driving cycles [2, 15]. As two main instantaneous optimization methods, the

equivalent consumption minimization strategy (ECMS) converts the electrical energy

consumption to equivalent fuel consumption upon which the calculation for optimal control

actions is based, and model predictive control (MPC), as explained by its name, makes

explicit use of a model and predicted information to solve the optimal control problem over

a given time horizon. Both strategies can be applied online due to relatively low

computational cost and easy implementation. Unlike ECMS that determines the control for

only one time instant, MPC obtains the optimal control action over a finite domain based

on online rolling optimization, which ensures good control and strong robustness [15]. As

a result, MPC has attracted much academic attention throughout the years, and it is widely

exploited on HEV in combination with a variety of strategies, including the PMP [16],

quadratic programming [17], nonlinear programming [18] or SDP [11, 19].

At the same time, the development of connected and automated vehicles (CAVs) in the

intelligent transportation system (ITS) has provided a huge possibility in improving fuel

economy and emission minimization as well as road safety. With the help of ITS and the

upcoming ubiquitous 5G network, CAV is expected to obtain real-time data seamlessly

from different sources including information via Vehicle-to-Infrastructure (V2I)

communication, such as road speed limit, traffic flow data and accident location, and

information via Vehicle-to-Vehicle (V2V) communication such as the position and

velocity of other vehicles [20–24], as shown in Fig. 1. 3. In addition, in comparison to

5

human-driven vehicles, automated vehicles are designed to reliably and accurately execute

real-time commands derived from control strategies [25]. Consequently, by combining the

technologies of HEVs and CAVs, vehicles can achieve even better performance in road

safety, energy management and emissions minimization. Current studies on connected and

automated hybrid electric vehicles (CAHEVs) are limited to the applications to certain

driving scenarios, such as regenerative braking scenario [26], car-following scenario [6]

and traffic signal anticipation scenario [27]. However, since a common day-to-day driving

trip can have complex driving conditions and consist of a combination of different driving

scenarios, it is not optimal and implementable to apply the results from one scenario to

some others.

Furthermore, the concept of eco-driving has been getting more and more attention

because the fuel consumption and exhaust emissions are also highly correlated to the

driving style of a vehicle [28]. Eco-driving, as the name explains, refers to an ecological

and economical driving style that can reduce carbon emissions and improve fuel economy

[29, 30]. The eco-driving method can involve techniques such as maintaining steady

vehicle speed, reducing the frequency of brake pedal use, driving in a high transmission

gear, shifting up in advance, and braking smoothly [31]. Conventionally, on an internal

combustion engine (ICE) vehicle controlled by human drivers, eco-driving techniques are

utilized through the application of Eco-Driving Assistance System (EDAS) or Eco-Driver

Feedback System (EDFS) that are able to assist human drivers by providing fuel-saving

information and eco indication and advice for achieving better driving behaviors [28, 31,

32]. Significant improvements have been observed by applying eco-driving techniques,

according to a research [30] based on the eco-driving studies from 1985 to 2011,

approximately 12% of fuel consumption was saved by implementing eco-driving systems

in comparison to the results by conventional driving styles. However, with the future

advancement of CAHEV technologies, the application of the eco-driving approach is

expected to be more challenging yet effective on a fully automated vehicle in a connected

traffic environment for the purpose of improving fuel economy and minimizing exhaust

emissions.

6

The motivation of this research is to propose and design a novel eco-driving control

strategy for connected and automated hybrid electric vehicles (CAHEVs) that is applicable

to trips involving mixed driving scenarios. The objectives of the proposed strategy are

improving energy management, reducing exhaust emissions and ensuring safe driving

requirements. This research can also build the gap between the study of energy

management strategy (EMS) applied to HEV and urban mobility regulation.

Fig. 1. 3 V2V and V2I communications for CAVs in ITS

1.2 Objectives and assumptions

The primary goal of this research is to design and develop an eco-driving strategy for

CAHEVs such that safe driving can be accomplished with improved fuel economy and

fewer exhaust emissions, under a combination of mixed eco-driving scenarios. The

proposed control strategy outputs the real-time commands at two levels. At the upper level,

7

the vehicular longitudinal acceleration is determined based on the real-time data received

from the communications of ITS. At the lower level, the split torque ratio between the

internal combustion engine (ICE) and the electric motor (EM) is optimized. Although the

lower level control strategy is designed specifically for HEVs, the vehicular longitudinal

control strategy at the higher level can be easily extended to vehicles of various types, such

as conventional internal combustion engine vehicles, plug-in hybrid electric vehicles,

electric vehicles, etc.

The detailed objectives of this research include:

(1) To model an HEV of the parallel hybrid configuration, which includes an internal

combustion engine (ICE), an electric motor (EM) and a battery pack.

(2) To build up the real-time data exchange based on Vehicle-to-Vehicle (V2V) and

Vehicle-to-Infrastructure (V2I) communications in a realistic traffic simulation

environment.

(3) To design a model predictive control (MPC) based automated eco-driving strategy

that optimizes three goals during driving 1) safe driving, 2) energy-saving, and 3)

emission reduction.

(4) To design a real-time classifier of driving scenarios for the studied vehicle in a

realistic traffic simulation environment. The driving scenarios are classified into

three cases 1) free driving scenario, 2) signal anticipation scenario and 3) car-

following scenario.

(5) To develop the cost function for the above three objectives. And to tune the

weighting factors among the proposed multi-objective optimization to obtain

reasonable and optimal results.

8

(6) To make comparative studies between the proposed MPC-based strategy and the

conventional rule-based strategy in a realistic traffic simulation environment

The following assumptions are made throughout the thesis

(1) The communication time via Vehicle-to-Vehicle (V2V) and Vehicle-to-

Infrastructure (V2I) is neglected.

(2) All the roads in the network are assumed to be flat road, i.e. road grade is zero.

(3) The distances covering the intersection regions are neglected in odometry

(4) The background traffic simulation is accident-free and collision-free

(5) Only the host vehicle is connected and automated vehicle (CAV), the background

vehicles are assumed to be connected vehicles (CVs) only

1.3 Contributions

To the knowledge of the author, the main contributions of this research are listed below:

(1) A bi-level MPC-based eco-driving for CAV is proposed to improve fuel economy,

reduce emissions based on driving safety.

(2) The proposed strategy can be adapted to trips with a combination of mixed driving

scenarios, namely, the free driving, the signal anticipation at the traffic signal

controlled intersections (TSCIs) and car-following scenarios.

9

(3) The system makes use of real-time information via Vehicle-to-Vehicle (V2V) and

Vehicle-to-Infrastructure (V2I) connections to exploit the potential of MPC in

realizing the objectives.

(4) The simulation results based on a realistic simulation map are presented and

compared to the rule-based strategy.

(5) The proposed strategy provides a promising opportunity to exploit the potential of

CAHEVs under realistic trips with all-inclusive driving scenarios.

1.4 Outline

The thesis is organized as follows:

• Chapter 1 Introduction introduces the challenges in the research area and the

latest technologies which the study is based on. Also, the motivation of this work

is presented, as well as the unique contributions from this study.

• Chapter 2 Literature Review reviews the existing energy management strategies

for HEVs, the eco-driving strategy for CAVs. A review of the blending of these

two strategies is also given, and the research gaps are identified and explained.

• Chapter 3 Model Development introduces the studied HEV model, including

the modelling of the internal combustion, electric motor, battery pack and the

longitudinal dynamic model. Followed by the introduction of the Vehicle-to-

Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) in the intelligent

transportation system (ITS)

10

• Chapter 4 Methodology describes the eco-driving problem to be solved, which

is then formulated in the framework of the MPC-based algorithm regarding the

objectives of driving safety, energy economy and emission. The benchmark eco-

driving strategy that is made up of a rule-based energy management strategy and

a realistic driver model is also introduced.

• Chapter 5 Results and Discussion provides the simulation details, and the results

run by the proposed MPC-based eco-driving strategy and the rule-based strategy

are shown and compared. Finally, the analysis based on the comparison results is

given.

• Chapter 6 Conclusion and Future Works concludes the current work and

suggest future works to be done.

11

Chapter 2. Literature Review

2.1 Introduction

So far, the control strategies for HEVs and CAVs have been extensively studied in the

literature. This chapter aims to provide a comprehensive review of the existing literature,

emphasizing mainly on the contributions to the control strategies for the connected and

automated parallel hybrid vehicles (CAHEVs). In the end, the research gaps are also

identified and explained.

The review will be initiated by the introduction to the research background, HEV

configurations and CAV technologies. Then, the existing literature on control strategies for

HEV and CAHEV will be reviewed and evaluated in depths, concerning their control

principles, advantages, drawbacks, limitations as well the contributions towards the

fulfillment of several optimization objectives, including the minimization in fuel

consumption and exhaust emission, the ability to maintain the battery SOC and the driving

safety-related issues, if applicable. Finally, the research gaps in the literature will be

identified and explained to pave the way for developing the proposed strategy in the

following chapters.

2.2 Hybrid electric vehicles

Hybrid electric vehicles refer to the type of vehicle that is equipped with a hybrid

powertrain consisting of not only the conventional fuel supplier but also the additional

electrical energy supplier from the onboard battery pack [2, 8]. For this reason, they are

considered as the transitional technologies for the automotive industry to migrate from

conventional internal combustion engine vehicles to pure electric vehicles. The hybrid

12

configuration design offers great flexibility in energy split control while ensuring the power

demand from the wheels [2].

Compared with the operation of conventional vehicles, the internal combustion engine

(ICE) on the HEVs can be assisted by the electric energy supplier when the engine

efficiency is relatively poor, and it can be fully decoupled from the drivetrain when there

is sufficient electric power to drive the vehicle. Besides, the regenerative braking can

restore the kinetic braking energy from deceleration to charge the battery pack, which is

otherwise wasted in the process of braking as for ICE vehicles [33]. As a result, a better

fuel economy and fewer emissions from the engine can be achieved.

One of the widely used methods to classify HEVs is based on the configurations of the

vehicle powertrain. They are classified as 1) Series Hybrid, 2) Parallel Hybrid, and 3)

Series-Parallel Hybrid.

2.2.1 Series Hybrids

The series hybrid configuration generally involves an internal combustion engine (ICE), a

generator, an electric motor (EM), an electrical energy storage system, as shown in Fig. 2.

1 [34]. It is often likened to the battery electric vehicle (BEV) configuration since there is

no direct mechanical connection between the internal combustion engine (ICE) and the

wheels, rather, the power provided from the ICE is converted to electrical power by the

generator to charge both the battery pack and power the EM to drive the wheels [2] It’s

worth noting that the regenerative braking is taken to charge the battery pack when the

vehicle is decelerating or driving downhill. During urban driving, the ICE is turned off,

and the electrical energy stored in the battery is transmitted to the EM to propel the vehicle.

During country driving, the ICE is turned on as there is a large amount of power required

from the wheels at high speed [35]. The advantage of this design is that the ICE can always

operate at its high-efficiency working region and reduce fuel consumption [10, 34]. This

benefit, however, is outweighed by the fact that there is a high requirement for the onboard

13

battery pack to be expensive and powerful with high energy density [2]. Series hybrid

configuration is often applied to heavy-duty commercial vehicles such as New Tesla Buses,

TEMSA Avenue Hybrid Buses, Mercedes Citaro Buses, and so on [2, 35].

Fig. 2. 1 The series hybrid powertrain

2.2.2 Parallel Hybrids

In the parallel hybrid configuration, the powertrain consists of an internal combustion

engine (ICE), an electric motor (EM), a battery pack as the electrical energy storage system,

as shown in Fig. 2. 2. It is named parallel because both the ICE and EM can work

independently or together to provide working torques to propel the vehicle [2, 34].

Different from the series hybrid configuration, the ICE of the parallel hybrid powertrain

has a mechanical connection to the wheels via a gearbox. The EM can work as an assistant

power supplier or the sole power supplier to the ICE depending on the power size of EM,

14

and it can work as a generator during the battery pack being charged by either the ICE or

the kinetic power from regenerative braking [2]. When the battery state of charge (SOC) is

high, the vehicle can be driven by the EM alone or by both ICE and EM together. When

the battery state of charge (SOC) is low, the redundant power output from the ICE can be

used to charge the battery pack. The advantage of this design is the flexibility of choosing

between the ICE and EM to propel the vehicle, which decouples the ICE from the

powertrain and enables it to work in the high-efficiency mode. However, with dual driving

machines, it often leads to a more complex control strategy and high computational cost

[35]. In addition, the working modes of parallel hybrid vehicles do no perform well in the

frequent stop-and-go condition as in urban driving. Examples of vehicles with hybrid

electric configurations include Honda Insight, Ford Escape Hybrid SUV and Lexus Hybrid

SUV [35].

Fig. 2. 2 The parallel hybrid powertrain

15

2.2.3 Series-Parallel Hybrids

The series-parallel hybrid configuration consists of an internal combustion engine (ICE),

motor/generator (M/G1), motor/generator (M/G2), inverter, battery pack, and power-split

device, as shown in Fig. 2. 3. As the name explains, the series-parallel hybrid combines the

advantages of both series hybrid design and parallel hybrid design while getting rid of their

disadvantages by integrating both the series and parallel energy path, as shown in Fig. 2. 3

[10, 36]. As a result, the powertrain component sizing problem can be solved by running

the parallel hybrid mode. On the other hand, the frequent stop-and-go condition, which is

unsuitable for parallel hybrid design, is overcome by the ability to have the battery charged

when the vehicle remains stopped [10]. However, in comparison to the aforementioned

configuration, the additional flexibility offered by the series-parallel hybrid configuration

adds more complexity in the powertrain structure and the control strategy that can take the

full use of its potential. Applications of the series-parallel hybrid configuration on

commercial vehicles include Toyota Prius, Lexus LS 600h and Nissan Tino [37].

16

Fig. 2. 3 The series-parallel hybrid powertrain

2.2.4 Driving Cycle

A driving cycle is a sequence of data points representing vehicle speed versus time [38,

39]. Besides, it can also include many other driving parameters concerning the resultant

vehicle performances, such as vehicle powertrain conditions, dynamometer settings,

transmission shift points, cold start conditions, and so on [40]. As a result, driving cycles

are often used as a way to evaluate and compare vehicle performances, e.g. fuel/energy

management and exhaust emissions, in a reproducible way [40].

Since the 1970s, great effort has been contributed to developing standard driving cycles

in many countries [41]. Among the well-established driving cycles, the widely used ones

include the New European Driving Cycle (NEDC) of Europe, Urban Dynamometer

17

Driving Schedule (UDDS) and EPA Federal Test Procedure (FTP-75) of USA and Japan

Cycle 08 (JC08) of Japan. Fig. 2. 4 illustrates the UDDS driving cycle.

Fig. 2. 4 The Urban Dynamometer Driving Schedule (UDDS)

2.3 Connected and automated vehicles (CAVs)

Connected and automated vehicles (CAV), as the name explains, is the combination of

connected vehicle (CV) and the automated vehicle (AV). In the technology of CV, road

vehicles are able to get real-time access to information from different resources via wireless

connections. With access to the information in the network, the driving performance can

thus be improved such as accident avoidance, road safety, time-saving, energy-saving and

emission reduction. The connections are classified into two main types, Vehicle-to-

Infrastructure (V2I) communication and Vehicle-to-Vehicle (V2V) communication. Via

Vehicle-to-Infrastructure (V2I) communication, a vehicle is able to access information

through the roadside infrastructures about the speed limit, weather forecast, school zone

and collision alert, traffic congestion location, and so on. With the help of Vehicle-to-

Vehicle (V2V) communication, vehicles in the same region mutually share vehicular data,

including vehicle type, position, speed, acceleration, intended action, route, and so on. As

18

a result, the comprehensive knowledge of the network gained in real-time can assist the

vehicle in a better decision-making process.

On the other hand, the automated vehicle (AV), which is fully controlled by vehicular

control systems, was designed to compensate for the inaccuracy and inconsistency in

human driving. For example, even with the knowledge of the road speed limit, the human

driver may exceed the limit by occasionally over pushing the pedal position. Also, limited

by the human attention span, a driver could make some fatal mistakes after an excessive

and continuous long driving. Therefore, the automated vehicle system is able to

consistently execute real-time driving commands accurately to the vehicle that ensures on-

road driving safety.

CAVs also offers great potential in improving the fuel economy and reduce the

pollutant emissions, as will be reviewed in Chapter 2.3.

2.3.1 Eco-Driving scenarios

Eco-driving stands for the type of control policy that the driver, either human or smart

vehicle, execute to minimize the fuel consumption and exhaust emissions. In this study,

the eco-driving is narrowed to the set of control commands by the automated vehicle

system to drive the connected and automated hybrid electric vehicles (CAHEVs) for

reducing the fuel consumption and pollutant emissions. Many factors are involved that

affect the eco-driving performance, including vehicle speed, acceleration and deceleration,

idling state, and so on. On the other hand, the optimal route choice belongs to the category

of eco-routing. Therefore, it is out of the scope of this study.

Some of the most common eco-driving scenarios can be presented in a non-exhaustive

list, which includes [25]:

19

1. Accelerating to a cruise speed: the vehicle begins with a given speed 𝑣𝑖 , and

continuously accelerates to a target speed 𝑣𝑓 (𝑣𝑓>𝑣𝑖)

2. Cruising: the vehicle keeps the speed 𝑣𝑖 constant to a target speed 𝑣𝑓=𝑣𝑖

3. Decelerating to a final speed: the vehicle continuously decelerates its speed 𝑣𝑖 to a

lower target speed 𝑣𝑓 (𝑣𝑓<𝑣𝑖)

4. Driving between stops: the vehicle begins with 𝑣𝑖=0 and ends the trip with 𝑣𝑓 = 0

5. TSCI approaching: the vehicle approaches a TSCI from a speed 𝑣𝑖 to a speed 𝑣𝑓 either

𝑣𝑓 = 0 when the traffic signal does not permit vehicle passing, or 𝑣𝑓>0 when the

traffic light allows vehicle passing

6. Urban trip: the vehicle drives in urban condition with frequent stop-and-go which can

involve 1~5

7. Highway trip: the vehicle drives in highway condition in a relatively high driving

speed with few to none stop, which can involve 1~4

8. Car following: the vehicle speed 𝑣𝑖 is constrained by the preceding vehicle speed 𝑣𝑓

since a safe inter-vehicle distance is meant to be kept

For the sake of designing an eco-driving system for all the above scenarios, the listed

ones are therefore classified into three main categories, as (I) free driving scenario: includes

case 1~4, (II) signal anticipation scenario: includes case 5, and (III) car-following scenario:

includes case 8. Case 6 and 7 are compounded cases that can be achieved by realizing their

component cases. In the rest of the study, the control strategies for connected and

automated hybrid electric vehicles (CAHEVs) will be proposed and designed in accordance

with each of the above-mentioned categories.

20

2.4 Literature review on control strategies for CAHEVs

2.4.1 Energy management strategies (EMSs) for HEVs

The main challenge in the powertrain operation of HEV comes from the optimal control of

power flow between the ICE and the EM to meet the driving demand while improving the

energy economy and reduce the exhaust emissions at the same time [10] Regardless of the

type of HEVs, adopting the right energy management strategies (EMSs) are crucial to the

realization of these goals [2].

The topic of EMS has been extensively studied with the development of HEV during

the past decades, and a great many strategies have been proposed. For the EMS, the inputs

usually include vehicle power demand, vehicle speed and acceleration, battery state of

charge (SOC) and sometimes the future traffic condition obtained from GPS or intelligent

transportation system, if applicable [2]. The output commands of control strategies decide

in what mode the vehicle should operate, for example, regenerative braking mode, ICE-

only mode, and so on [2]. Until now, the existing EMS can be classified into two main

categories: 1) rule-based control and 2) optimization-based control, as shown in Fig. 2. 5.

Prior to the introduction to different types of EMSs, the HEV operation modes will be

described.

21

Fig. 2. 5 The classifications of control strategies for HEV

22

2.4.1.1 HEV working modes

The control action in a parallel hybrid electric powertrain mainly concerns how to

determine the best split ratio between the mechanical power provided by the ICE and the

electrical power provided by EM in a way that minimizes the optimization goals. The

power paths, along with the vehicle modes, are shown in Fig. 2. 6 and Table. 2. 1, and the

split ratio to be determined is defined as the ratio of the electrical power 𝑃𝐸𝑀(𝑡) to the total

required power 𝑃𝑟𝑒𝑞(𝑡) as,

Table. 2. 1 HEV working modes

Operation mode 𝑢(𝑡) Clutch

Engine only 0 Open

Electric only 1 Closed

Hybrid / Electric assist (0,1) Open

Battery charging (-∞, 0) Open

Regenerative Braking (-∞, 0) Closed

Among the above working modes, the engine only mode is when the sole vehicle power

depends on the output power of ICE that helps ICE to remain in its high-efficiency region.

The electric-only mode refers to the mode when the vehicle is propelled by the electrical

energy supplied by the EM. It is on when the ICE operates in low efficiencies, such as at

low vehicle speed. The hybrid / electric assist is on either when the required power from

the wheels is higher than the power provided by the ICE, or to make the ICE work at high

torque region. When the battery SOC is lower than the preset limit, the vehicle switches to

battery charging mode to charge the battery using excessive engine power. When the

𝑢(𝑡) =𝑃𝐸𝑀(𝑡)

𝑃𝑟𝑒𝑞(𝑡) (1. 1)

23

vehicle is decelerating, the EM is able to work as a generator to absorb the braking kinetics

to charge the battery.

Fig. 2. 6 The vehicle working modes, (a) electric only, (b) hybrid / electric assist, (c)

battery charging, (d) regenerative braking, (e) engine only

24

2.4.1.2 HEV rule-based control strategies

The rule-based control strategies, as the name indicates, make use of simple rules derived

from human expertise (engineering knowledge), mathematical models, intuition with the a

priori information of a driving cycle, to control the power flow in the hybrid powertrain

[35, 37]. They are the most widely used strategy to implement in real-time for an HEV [2].

The advantage of rule-based control strategies is their low computational cost and

simplicity in real-time execution [10]. The drawbacks, however, are that the set of pre-

defined rules is only optimized for a certain driving cycle or driving condition. Therefore

it produces results that are non-optimal or sometimes far from optimal unless the

knowledge of real-time driving cycle is fully apprehended [10] The rule-based strategies

can be further categorized as deterministic-based strategies and fuzzy-based strategies.

In deterministic-based strategies, a pre-defined lookup table that has a set of fixed rules

is used to deliver real-time control to the vehicle [37]. For example, Kim et al. [42]

proposed a deterministic rule-based strategy for HEV with the continuously variable

transmission (CVT) that derived the hybrid optimal operation line (HOOL) based on a set

of principles for converting battery power to equivalent fuel consumption. As a result,

optimal control variables such as gear ratio, engine torque and motor torque can be obtained

online, and fuel consumption can be reduced. Similarly, in order to make a parallel HEV

work properly according to the power demand and battery SOC, the power follower

(baseline) control strategy [43] controls the engine on/off state based on a set of pre-defined

rules. Also, Jalil et al. [44] designed a thermostat control strategy for a series HEV that

applied a simple rule to maintain the battery SOC in the upper and lower bounds by turning

the engine on/off. The results were compared to another rule-based strategy to show the

improvement in fuel economy. Electric-assist control strategy (EACS) [45] probably is the

most successfully applied and widely used deterministic rule-based strategy for HEV in

which the engine works as the main power source while the motor serves to assist the

engine when the required demand is higher than the engine can provide, or when the engine

works in low-efficiency region [2]. The maintenance of the SOC level is also considered

25

in the rules. However, a large number of parameters in EACS increase the sensitivity to

driving patterns and result in less robustness.

Though the above mentioned deterministic rule-based strategies show good results in

certain driving condition and profile, however, a huge amount of time will be needed to

tune and calibrate the strategies to work for other driving conditions, thus are less adaptive

and robust to complex driving conditions such as real-world traffic scenario. In response

to the above disadvantages of deterministic rule-based strategies, fuzzy rule-based

strategies were designed to tackle those issues.

Basically, fuzzy-based control strategies are derived from conventional deterministic

rule-based strategies. However, unlike the deterministic rules as in previous methods, in a

fuzzy rule-based controller, the linguistic commands are converted into multivalued logical

rules using fuzzy set theory. Consequently, they are more robust and tolerant of imprecise

measurements and system variations and more adaptive to various driving conditions [46].

For example, in [47], a fuzzy rule-based strategy was proposed for a parallel HEV. A

dynamo test was performed to set up the fuzzy relationship between the inputs of

acceleration pedal stroke and motor speed, and the output motor torque command. The

results show the robustness to disturbances and a 20% reduction in NOx emissions with

balanced battery level. Moreover, as the objectives of minimizing fuel consumption and

emissions are often conflicting, for instance, a decrease in fuel consumption would lead to

an increase in exhaust emissions, an adaptive fuzzy rule-based strategy [48] as designed to

make a trade-off between the objectives of fuel and emissions by using the weighted sum

approach. With the relative weights assigned, a significant reduction was found in

emissions at the expense of a negligible increase in fuel consumption. In addition, unlike

the above fuzzy rule-based strategies that aim to achieve local best solutions, a global best

solution can be obtained by obtaining the knowledge of the trip in advance. Therefore,

Ichikawa et al. [49] proposed a predictive fuzzy rule-based strategy that is able to predict

the future driving pattern from the historical data. Based on the predicted driving pattern,

a set of fuzzy rules were set up to execute better engine commands for an HEV.

26

Typically, the rule-based strategies are easy to implement, hence widely used in real-

time control. Due to the fact that they are derived from heuristic rules and human expertise

rather than an optimization approach based on numerical models, they often produce near-

optimal results. The summary of rule-based strategies is listed in Table. 2. 2.

Table. 2. 2 Summary on rule-based energy management strategies for HEV

Strategy Category Online/Offline Descriptions

HOOL-based [42] Deterministic

rule-based

Online ▪ The hybrid optimal operation line (HOOL)

derived from a set of pre-defined rules

▪ Tuned for specific driving conditions

Power follower

[43]

Deterministic

rule-based

Online ▪ Engine power state controlled by a set of

rules


Thermostat [44] Deterministic

rule-based

Online ▪ Engine controlled by a set of rules for

balancing battery SOC


EACS [45] Deterministic

rule-based

Online ▪ Most widely used deterministic rule-based

strategy for HEV

▪ Motor assisting engine based on a set of

rules with battery SOC considered


Conventional fuzzy

[47]

Fuzzy rule-

based

Online ▪ Motor operation determined by a set of

fuzzy rules from dynamo test

▪ More adaptive and robust than

deterministic rules

Adaptive fuzzy

[48]

Fuzzy rule-

based

Online ▪ Weighted-sum optimization based on

adaptive fuzzy rules


deterministic rules

Predictive fuzzy

[49]

Fuzzy rule-

based

Online ▪ A set of fuzzy rules for predicting future

driving patterns based on historical data


deterministic rules

27

2.4.1.3 HEV optimization-based control strategies

In optimization-based strategies, the control system adopts an analytical or numerical

approach to minimize a cost function that can have either single or multiple objectives [37].

They can ensure a higher degree of optimality than rule-based strategies. However, the

calculation usually imposes a heavy burden on the computer. They can be further divided

into global optimization and real-time optimization.

In global optimization, the entire knowledge of driving conditions is required a priori

to derive the global optimal control policies for HEV. For example, dynamic programming

(DP), which is derived from Bellman optimality equations, is often used as a benchmark

method for other strategies because it can guarantee global optimal results once the

knowledge of driving pattern is given. In the study [50], dynamic programming (DP) was

applied on a hybrid electric truck to find the global optimal control policies, including the

gear shifts and power split between the engine and motor while keeping the battery in a

balanced SOC. The optimal results show a considerable amount of emissions were reduced

in the sacrifice of a small amount of fuel increase. However, the results calculated by the

dynamic programming are only optimized for a specific driving cycle, that is to say, the

results are not guaranteed to be global optimal when the results are applied in other driving

conditions. As a remedy, stochastic dynamic programming (SDP) was utilized by Lin et al.

[51] on a parallel HEV to obtain optimal instantaneous engine output power. In this SDP,

on the basis of a few pre-defined typical driving cycles, the vehicle power demand is

modelled as a random Markov process in which the current state variables can be derived

by a state transition probability solely from the previous state instead of from the historical

states. As a result, the optimized control policy is in the form of stationary full-state

feedback over a set of preset random driving cycles in an average sense [2] and can be

applied to different driving cycles. Though more adaptive than DP, the optimality of SDP

is highly concerned with the similarity between the pre-defined driving cycle and the tested

driving conditions, including the type of roads, the impact of weather, mixed traffic

congestion, and so on. Moreover, DP and SDP suffer from high computational cost due to

28

the intrinsic complexity of Bellman equations. Genetic algorithm (GA) is another approach

to achieve global optimality, which is inspired by Darwin’s theory of evolution. GA begins

with a group of chromosomes representing the preliminary solutions. The chromosomes

can perform mutation and crossover in order to produce the offspring candidate for the

optimal solution, and this process repeats itself until the terminal condition is reached [2].

It is useful for solving constrained nonlinear optimization problems [46]. For example, GA

was adopted in [52] for a series HEV. In this study, the problem of fuel economy and

emission reduction is formulated as a multi-objective optimization and is solved by GA.

The results demonstrate the effectiveness of GA in solving constrained multi-objective

optimization. Nevertheless, due to the time-consuming computation spent on mutation,

crossover and selection, GA is not readily implementable in real-time. Another approach

originated from natural phenomenon is particle swarm optimization (PSO). It was inspired

by the social behaviour of bird-flocking in which the particles that are the representation

of birds are searching around a solution space for improving better solutions over the steps

[2]. In [53], for example, Huang et al. adopted PSO on the multilevel hierarchical control

to minimize the fuel consumption for a parallel HEV. The results showed an improved fuel

economy was achieved in comparison to a built-in control strategy in the PNGV system

analysis toolkit (PSAT). Like previously mentioned global optimization strategies, PSO

also suffers from high computational cost and is not suited to be used online.

On the other hand, the real-time optimization strategies intend to minimize

the instantaneous cost function in a short domain. This is accomplished by breaking down

the global optimization into a sequence of local optimization. Consequently, the total time

cost can be greatly shortened [2]. Despite producing sub-optimal results when compared

with global optimization approaches, real-time optimization strategies have received great

attention from researchers in the field of HEV control [2]. For instance, Pontryagin's

minimum principle (PMP) is among the most popular real-time optimization strategies for

HEV power management. PMP can be used under the assumption that the optimal solution

can be found through the instantaneous minimization of a Hamiltonian function over a

given driving cycle [2, 54]. For example, in the study of [55], PMP was employed to derive

29

the control law for a plug-in HEV. The results demonstrate that by an acceptable

approximation of the co-state from the past driving patterns, the produced results can be

near-optimal in comparison to the global optimal solutions produced by DP. The drawback

of PMP lies in that the co-state as the only variable is highly related to the selected driving

patterns. As a result, the optimally tuned PMP that works for a family of driving patterns

can be suboptimal in other driving patterns. While in equivalent consumption minimization

strategy (ECMS), the solving Hamiltonian function is simplified to the calibration of the

equivalent factor 𝜆 that balances the use between fuel and electrical energy. It is proposed

by Paganelli et al. [56] as a way to achieve an optimal solution by minimizing the

instantaneous weighted sum cost function consisting of the actual fuel consumption, and

the equivalent fuel consumption from the electrical energy [11]. In the study by Sciarretta

et al. [57], the cost function that adds up the instantaneous fuel consumption and the

equivalent fuel consumption based on the electrical energy is minimized at each time step,

and a fuel reduction up to 50% was found when compared with other conventional

algorithms. The problem with ECMS is that the calibration of the equivalent factor 𝜆

requires full knowledge of the driving cycle or future driving prediction in the optimization

[58]. Whereas the actual driving pattern often differs from the pre-defined one in ECMS,

the results will be compromised and thus is not optimal. Model predictive control (MPC)

is another type of strategy that can be implemented in real-time. In MPC, the optimal

control problem in the finite control horizon is solved at each time instant, and the optimal

set of control actions is obtained as the first element of the optimal control policy in the

control horizon. To solve optimization problems, MPC can be matched to a variety of

strategies including the PMP [16], quadratic programming [17], nonlinear programming

[18],or SDP [11, 19]. In recent years, different types of MPC have been widely used in

HEVs due to its rapid computation, capability in dealing with constrained multivariable

problems and potential for solving optimization problems as a receding horizon control

strategy [11]. For example, Sun et al. [59] developed a bi-level MPC-based energy

management strategy for plug-in HEV that utilizes the real-time traffic data to control the

battery SOC and reduce fuel consumption simultaneously. The results showed a 94%~96%

fuel optimality was achieved in comparison to the results by DP. Similarly, in the study

30

[27], a bi-level MPC was applied to an HEV to achieve eco-driving in an urban driving

condition. In this control strategy, the optimal vehicle velocity is determined at the upper

level, while the torque split ratio and gear shift are optimized at the lower level. The results

show an improvement in fuel economy. The summary of optimization-based strategies is

listed in Table. 2. 3.

Table. 2. 3 Summary on optimization-based energy management strategies for HEV

Strategy Category Online/Offline Descriptions

DP [50] Global

optimization

Offline ▪ Often used as a benchmark strategy

▪ Required knowledge of driving conditions

▪ Optimized for the specific driving cycle(s)

▪ High computational cost

SDP [51] Global

optimization

Offline ▪ Based on a state transition probability

obtained by driving cycles

▪ More adaptive than DP


GA [52] Global

optimization

Offline ▪ Inspired by Darwin’s theory of evolution

▪ Complexity growing with solution spaces


PSO [53] Global

optimization

Offline ▪ Inspired by natural phenomenon


PMP [55] Real-time

optimization

Online ▪ Among the most popular strategies

▪ Tuned for specific driving cycles

▪ For real-time implementation

ECMS [57] Real-time

optimization

Online ▪ Simplified version of PMP

▪ Tuned for specific driving cycles


MPC [27, 59] Real-time

optimization

Online ▪ Widely used strategy on HEVs

▪ Capable of dealing with constrained

multivariable problems

▪ A receding horizon control strategy


31

2.4.2 Control strategies for CAHEVs

In addition to the potentials of improving the powertrain working efficiency enabled by

various EMSs of HEVs, the recent emergence of connected and autonomous vehicles

(CAVs) in the context of Intelligent Transportation Systems (ITS) have offered an

unprecedented opportunity for energy saving and emission reduction in automobiles. So

far, some studies have been done concerning the control strategies for connected and

automated hybrid electric vehicles (CAHEVs). The following reviews will be done for

each of the categories listed in Chapter 2.3.1.

For the scenario of free driving, for example, in [60], the integration of V2I data into

the MPC-based EMS enables the system to obtain an optimal driving profile, and reduced

vehicle emissions for a hybrid electric bus over a fixed route with knowledge obtained a

priori. Also, in the study [61], based on a reference driving cycle, the vehicle speed

trajectory was generated offline by dynamic programming (DP) to realize the goal of fuel

economy, and the results indicate the effectiveness of applying eco-driving to vehicles with

HEV configuration. Similarly, in [62], given a series of target speed points, the torques and

speeds of ICE and EM are determined by an MPC-based strategy to control the vehicle

speed to improve the driving performance. In [26], Wu et al. designed a hierarchical control

strategy with the consideration of the battery ageing problem for HEV in the case of

regenerative braking. During the braking process, an upper-level controller is able to

control the decelerating speed, which is sent to the lower level to optimize the power path

in the vehicle powertrain system. The simulation was carried out on a hybrid electric bus,

and the results were superior to the one computed by a logic-based strategy.

Considering the scenario that involves a traffic signal system, for example, in [27], Guo

et al. proposed a bi-level MPC-based strategy. At the upper level, the optimal speed profile

is achieved based on the traffic signal information. At the lower level, the MPC strategy is

utilized for energy management problem in HEV. In the study [63], the author proposed

an MPC-based eco-driving strategy for HEV that is capable of optimizing the speed profile

in the scenario of TSCI and sloped roads. The results were compared to the case of the

32

human driver. It was demonstrated that by equipping this speed control strategy, the speed

profile along with the fuel economy was improved greatly.

In order to achieve eco-driving during the car-following scenario, for example, in the

study [6], a two-level strategy is devised for a multi-objective problem. The upper level of

the system aims at keeping the host vehicle within a safe distance while following the

preceding vehicle. The lower level applies the cost derived from ECMS on the powertrain

to help the engine work in high efficiency and keep a balance between the fuel energy and

electrical energy use. Similarly, Shieh et al. [64] designed an online eco-driving strategy

for HEV to achieve the objectives of fuel-saving and driving performance. This strategy

particularly deals with the pulse and glide (PnG) operation in the car-following scenarios.

As to the eco-driving in a combination of different scenarios, for example, in [65],

based on the real-time data obtained from ITS communications, Zhao et al. proposed an

integrated eco-driving control strategy for intelligent hybrid vehicles (IHV) for two

scenarios 1) adaptive cruise control (ACC) during car-following case and 2) regenerative

braking control (RBC) during vehicle deceleration. There are three key components in this

strategy, namely, the ACC to keep the safe distance with the preceding vehicle, the RBC

to determine the optimal powertrain torque distribution, and the coordinated control

strategy for both ACC and RBC. The simulation was performed on a straight single-lane

with no traffic lights, and the result demonstrates that 5.9% in the energy cycling efficiency

can be achieved by this algorithm when compared with a rule-based strategy. In work [66],

in order to achieve optimal fuel economy, maintain SOC level and execute safe driving,

the author developed a multi-level eco-driving algorithm for HEV that includes two stages.

At the first stage, a global optimal speed profile is generated offline based on the historical

data on a chosen route with traffic conditions indicated by the distance between the host

and the preceding vehicle. At the second stage, the host vehicle adapts its speed to the

simulated traffic on a short horizon. Though the combination of these two stages produces

an effective eco-driving method for HEV, however, throughout the trip, the host vehicle

only interacts with the single preceding vehicle, and the measurement in traffic congestion

33

level is embodied in the distance between those two vehicles which makes it oversimplified

and impractical in real driving situations.

2.4.3 Existing research gaps in eco-driving for CAHEVs

As reviewed thus far, most of the research efforts have addressed the eco-driving problem

of CAHEVs in limited driving scenarios, for example, regenerative braking control as

studied in [6], car-following control as studied in [6, 65], speed regulation based on traffic

signals as studied in [45]. However, to address the eco-driving problem on an actual day-

to-day trip in ITS, a vehicle can encounter mixed driving scenarios [67]. For instance, a

vehicle may accelerate from a complete stop until its speed reaches the road speed limit,

then the vehicle anticipates the upcoming traffic signal when it is approaching an

intersection, or a vehicle may follow a preceding vehicle until a right turn comes that leads

it to another vehicle to follow [67]. As a result, regarding the application of CAHEVs in

ITS, an eco-driving strategy that deals with a certain driving scenario may be far from

optimality or adaptability when it is applied to another driving scenario. Therefore, it is

desirable to design an eco-driving algorithm for CAHEVs to apply in ITS that can deal

with a combination of the most common driving scenarios. In this study, an actual day-to-

day driving can be broken down into three basic scenarios depending on most common

driving scenarios in Chapter 2.3.1, as 1) free driving: when the vehicle can drive with no

obstacle ahead to reach the road speed limit; 2) signal anticipation: when the vehicle makes

safe decisions in approaching a traffic signal-controlled intersection; 3) car-following:

when the vehicle follows a preceding vehicle in the same lane [67].

In addition, as two of the main goals in eco-driving problem, the optimization of fuel

economy and the reduction of engine-out pollutant emissions are usually achieved

separately in current studies as they are often conflicting, e.g. when the fuel is decreased,

the exhaust emissions may go up and vice versa [45, 67]. However, with the approach of

weighted-sum optimization, they can be achieved simultaneously by making a tradeoff in

34

the eco-driving problem [45]. So far, no work has been done regarding the application of

CAHEVs to simultaneously improve energy economy and exhaust emissions reduction

with driving safety ensured on a day-to-day ITS trip that involves mixed driving scenarios.

To explore the eco-driving potential of CAHEVs and address the limitations of the

aforementioned literature, an online eco-driving control strategy for CAHEV is proposed

and designed in this study. The objectives to consider are fuel economy, exhaust emission

reduction and driving safety. To achieve the above goals, a bi-level MPC-based approach

is taken to solve the multi-objective optimization problem. At the lower level, the control

strategy aims to determine the optimal power distribution between ICE and EM, while at

the upper level, the vehicle speed profile is fully controlled and optimized by the intelligent

system.

2.5 Summary

In this chapter, the background of HEV and CAV technologies were introduced first, and

then, a literature review was done on the energy management strategies (EMS) for HEV.

A variety of EMSs were analyzed concerning their principles, advantages and limitations,

and contributions. Next, a number of control strategies that are applied to CAHEV were

assessed. The existing research gap was pointed out at the end.

35

Chapter 3. Model Development

3.1 Modelling of Hybrid Electric Vehicles (HEVs)

In this chapter, the host vehicle configuration is modelled first, followed by the modelling

of the intelligent transportation system (ITS) that provides real-time Vehicle-to-Vehicle

(V2V) and Vehicle-to-Infrastructure (V2I) data to help the host vehicle derive optimal

strategy for eco-driving scenarios.

3.1.1 HEV Configuration

This study considers a single-shaft hybrid electric vehicle (HEV) for the host vehicle,

which is equipped with a 63kW/1.9L internal combustion engine (ICE), a 49 kW electric

motor (EM), an automated mechanical transmission (AMT) and a battery pack, as shown

in Fig. 3. 1. The parallel power flow paths include the mechanical power path that transfers

the power supplied by the fuel tank to the ICE, then goes through the EM and drives the

wheels and the electrical power path that brings the electrical power from the battery pack

to the EM to propel the vehicle. The configuration of the host HEV is illustrated as follows.

36

Fig. 3. 1 Single-shaft pre-transmission HEV Powertrain

The powertrain is equipped with an internal combustion engine (ICE), a clutch, a

battery pack, an electric motor (EM), an inverter and an AMT transmission box. The

parallel power flow paths include the mechanical power path that transfers the power

supplied by the fuel tank to the ICE, then goes through the EM and drives the wheels and

the electrical power path which brings the electrical power from the battery pack to the EM

to propel the vehicle. The component specifications are listed in Table. 3. 1, which is

acquired from ADVISOR.

37

Table. 3. 1 The studied HEV specifications [68]

Components Specification

Architecture

Vehicle Length

Vehicle Mass (m)

Frontal Area (A)

Parallel hybrid electric

4.3 m

1751 kg

2.66 m2

Engine Saturn 1.9L (63kW) SOHC SI Engine

Motor Honda 49kW Permanent Magnet Motor

Battery 13Ah, 120 cells, ESS_NIMH6

Transmission AMT, Gear ratio: [3.79, 2.17, 1.41, 1, 0.86]

Final Drive 4.55

Friction Coefficient (μ) 0.0150

Aerodynamic Coefficient (𝑪𝒅) 0.30

Air Density (𝝆𝒂𝒊𝒓) 1.20 𝑘𝑔/𝑚3

The clutch works as an ON-OFF switch of the ICE, and it can either be engaged or

disengaged, denoted as 𝑆𝐼𝐶𝐸 = [0,1] [58]. When the clutch is engaged, 𝑆𝐼𝐶𝐸 = 1, the ICE

is connected to the powertrain thus can provide mechanical power. When the clutch is

disengaged, 𝑆𝐼𝐶𝐸 = 0, the ICE is decoupled from the powertrain while the electric motor

is the only machine that drives the vehicle.

The inverter can convert between the direct current (DC) power, which is the electricity

stored in the battery pack, and the alternating current (AC) power, which is generated by

the electric motor [69].

3.1.1.1 Internal combustion engine (ICE) model

ICE is the main mechanical power supplier of an HEV. In propelling the vehicle, the

ICE outputs the mechanical power at the sacrifice of the fuel supplied by the fuel tank, at

the same time, the exhaust emissions (𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥 ) are emitted into the air. The fuel

38

consumption rate of the ICE can be modelled as a function of the instantaneous engine

torque 𝑇𝐼𝐶𝐸 and angular speed 𝜔𝐼𝐶𝐸, as follows [6],

where 𝐵𝑆𝐹𝐶 stands for Brake Specific Fuel Consumption, and it is often in the unit of

𝑔/𝑘𝑊ℎ. The BSFC map of the studied Saturn 1.9L (63kW) SOHC SI Engine is illustrated

in Fig. 3. 2 (a).

The instantaneous fuel consumption per second of the ICE can be derived, as follows

[6],

and it is in the unit of 𝑔/𝑠. It is worth mentioning that the ICE can not operate below the

idle speed.

Similarly, the instantaneous ICE exhaust emissions (𝐻𝐶,𝐶𝑂, 𝑁𝑂𝑥) can be modelled

concerning the instantaneous engine torque 𝑇𝐼𝐶𝐸 and angular speed 𝜔𝐼𝐶𝐸, as follows [6],

where �̇�ℎ𝑐 , �̇�𝑐𝑜 and �̇�𝑛𝑜𝑥 are the instantaneous emissions of 𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥 in the unit of

𝑔/𝑠. The studied Saturn 1.9L (63kW) SOHC SI engine-out emission efficiency maps for

𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥 are displayed in Fig. 3. 2 (b), (c), (d), respectively.

𝐵𝑆𝐹𝐶 = 𝑓𝐼𝐶𝐸(𝑇𝐼𝐶𝐸 , 𝜔𝐼𝐶𝐸) (3. 1)

�̇�𝑓𝑢𝑒𝑙 = 𝑇𝐼𝐶𝐸 ∙ 𝜔𝐼𝐶𝐸 ∙ 𝐵𝑆𝐹𝐶

3.6 × 106 (3. 2)

{

�̇�ℎ𝑐 =

𝑇𝐼𝐶𝐸 ∙ 𝜔𝐼𝐶𝐸 ∙ 𝑓ℎ𝑐(𝑇𝐼𝐶𝐸 , 𝜔𝐼𝐶𝐸)

3.6 × 106

�̇�𝑐𝑜 =𝑇𝐼𝐶𝐸 ∙ 𝜔𝐼𝐶𝐸 ∙ 𝑓𝑐𝑜(𝑇𝐼𝐶𝐸 , 𝜔𝐼𝐶𝐸)

3.6 × 106

�̇�𝑛𝑜𝑥 =𝑇𝐼𝐶𝐸 ∙ 𝜔𝐼𝐶𝐸 ∙ 𝑓𝑛𝑜𝑥(𝑇𝐼𝐶𝐸 , 𝜔𝐼𝐶𝐸)

3.6 × 106

(3. 3)

39

Fig. 3. 2 Operation points of Saturn 1.9L (63kW) SOHC SI Engine on fuel economy

and engine-out emissions contour maps (a) Fuel economy (b) HC emission (c) CO

emissions and (d) NOx emission

3.1.1.2 Electric motor (EM) model

The EM is the electrical power supplier of the studied vehicle. Working as an electric motor,

it takes power from the battery pack and output to drive the vehicle. Working as a generator,

on the other hand, it absorbs the power sent from the ICE either from redundant engine

power or regenerative braking power, to charge the battery. The motor efficiency can be

(a) (b)

(c) (d)

40

modelled as a function of the instantaneous EM torque 𝑇𝐸𝑀 and angular speed 𝜔𝐸𝑀, as

follows [6],

The power of the EM can then be written as follows [6],

where 𝛼 equals to 1 when the EM works as a generator, and it equals to -1 when the EM

works as an electric motor.

3.1.1.3 Battery pack model

The battery pack is modelled using the equivalent circuit model with an open circuit

voltage equipped with an internal resistance [70], as shown in Fig. 3. 3. The battery power

that supplies the EM can be written as follows [59],

where 𝑉𝑜𝑐 , 𝐼𝑏𝑎𝑡𝑡 , 𝑅𝑏𝑎𝑡𝑡 are the open-circuit voltage, current in the circuit, and internal

resistance, respectively.

Fig. 3. 3 Equivalent circuit model for battery pack

𝜂𝐸𝑀 = 𝑓𝐸𝑀(𝑇𝐸𝑀 , 𝜔𝐸𝑀) (3. 4)

𝑃𝐸𝑀 = 𝑇𝐸𝑀 ∙ 𝜔𝐸𝑀 ∙ 𝜂𝐸𝑀𝛼 (3. 5)

𝑃𝑏𝑎𝑡𝑡 = 𝑉𝑜𝑐𝐼𝑏𝑎𝑡𝑡 − 𝐼𝑏𝑎𝑡𝑡2 𝑅𝑏𝑎𝑡𝑡

(3. 6)

41

The rate of the state of charge (SOC) that shows the remaining battery energy is

modelled as follows,

where 𝑄𝑏𝑎𝑡𝑡 is the maximum battery capacity.

Combining Eq. (3. 6) and Eq. (3. 7) results in the following equation,

The power exchange between the battery pack and the EM is represented as follows,

where 𝑇𝐸𝑀 and 𝜔𝐸𝑀 are the EM torque and its rotational speed. When 𝑇𝐸𝑀 is positive, the

battery power 𝑃𝑏𝑎𝑡𝑡 is positive, the battery pack is discharging. When 𝑇𝐸𝑀 is negative, the

battery power 𝑃𝑏𝑎𝑡𝑡 is negative, the battery pack is charging.

𝑆𝑂𝐶̇ = −𝐼𝑏𝑎𝑡𝑡𝑄𝑏𝑎𝑡𝑡

(3. 7)

𝑆𝑂𝐶̇ = −𝑉𝑜𝑐 −√𝑉𝑜𝑐

2 − 4𝑃𝑏𝑎𝑡𝑡𝑅𝑏𝑎𝑡𝑡

2𝑄𝑏𝑎𝑡𝑡𝑅𝑏𝑎𝑡𝑡

(3. 8)

𝑃𝑏𝑎𝑡𝑡 = 𝑇𝐸𝑀 ∙ 𝜔𝐸𝑀

(3. 9)

42

3.1.1.4 Vehicle Dynamics

Fig. 3. 4 Longitudinal vehicle dynamics

The required power for the vehicle is depicted in Fig. 3. 4 and given by the following

equation [71, 72],

where the 𝐹𝑟 , 𝐹𝑔, 𝐹𝑎, 𝐹𝑖 are the rolling resistance, gradient resistance, aerodynamic drag

force and acceleration resistance, respectively. 𝑚, 𝑣, and 𝛼 are the vehicle mass, vehicle

velocity and road slope angle, respectively. The road slope 𝛼 is set to zero. 𝑚𝑒 is the

effective mass including the vehicle gravimetric mass and rotational inertias of components

[73]. It is estimated to be 1.08 times the vehicle gravimetric mass [72, 73].

{

𝑃𝑟𝑒𝑞 = (𝐹𝑟 + 𝐹𝑔 + 𝐹𝑎 + 𝐹𝑖) ∙ 𝑣

𝐹𝑟 = 𝑚𝑔𝑓𝑟 cos 𝛼 𝐹𝑔 = 𝑚𝑔 sin 𝛼

𝐹𝑎 = 0.5𝜌A𝐶𝑑𝑣2

𝐹𝑖 = 𝑚𝑒�̇�𝑚𝑒 = 1.08𝑚

(3. 10)

43

3.2 Intelligent transportation system (ITS)

In order to optimize the control strategy for the host CAHEV in ITS, real-time traffic data

is obtained via Vehicle-to-Vehicle (V2V) communication and Vehicle-to-Infrastructure

(V2I) communications. As the ITS communication technology is not the main concern in

this study, the delay in data transmission is omitted in both V2V and V2I.

3.2.1 Vehicle-to-Vehicle (V2V)

With V2V connectivity, a vehicle can get the real-time information of the neighbouring

vehicles to make better control decisions, for example, in the car-following scenario, the

real-time data of the preceding vehicle enables the host vehicle to keep the safe distance

without the potential of collision. The V2V data used in this study includes the distance to

the preceding vehicle in the same lane, ∆𝑠, and the speed of the preceding vehicle, 𝑣𝑙𝑒𝑎𝑑,

as listed in the following table.

Table. 3. 2 V2V data for this study

Data Variable Definition Sample Value

∆𝑠 (m) The distance to the closet vehicle ahead i

n the same lane towards the destination 25.2m

𝑣𝑙𝑒𝑎𝑑 (m/s) The speed of the ahead vehicle in the sa

me lane

13.8m/s

44

3.2.2 Vehicle-to-Infrastructure (V2I)

In the urban network, the traffic situation is crucial in deriving an effective and safe vehicle

control strategy, and this is made possible via V2I communication. In the V2I

communication, the V2I data is sent from the roadside units to the host vehicle about the

requested real-time environment and traffic information such as the maximum lane speed,

the traffic light, the distance to the intersection, and so on. In this study, the V2I data used

includes the distance to the next intersection, ∆𝑑, the traffic signal of the upcoming TSCI,

𝑆𝑟𝑦𝑔, the remaining time for the current traffic signal phase, ∆𝑡𝑇𝐿, the speed limit for the

lane where the host vehicle is running on, 𝑣𝑙𝑎𝑛𝑒, as listed in the following table.

Table. 3. 3 V2I data used for this study


∆𝑑 (m) The distance to the upcoming TSCI 121.5m

𝑆𝑟𝑦𝑔 The traffic signal of the upcoming TSCI ‘g’

∆𝑡𝑇𝑆 (s) The remaining time for the current traffic signal phase 23s

𝑣𝑙𝑎𝑛𝑒 (m/s) The speed limit for the lane where the host vehicle is running on 27.8m/s

The traffic signal phases are defined as follows,

45

Table. 3. 4 Listed traffic signal phases

Phase Meaning

g Green light signal, vehicles are allowed to pass

y Yellow light signal, vehicles must decelerate

r Red light signal, vehicles must stop

3.3 Summary

In this chapter, the studied HEV powertrain has been modelled consisting of the internal

combustion engine (ICE), the electric motor (EM) and the battery pack. Next, the vehicle

dynamics have been built up to compute the required vehicle power demand at the wheels.

Finally, the intelligent transportation system (ITS) concerning the data communications via

Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) have been introduced,

which lays the foundation for the design of the proposed eco-driving strategy in the

following chapters.

46

Chapter 4. Methodology

4.1 Problem Formulation

The control objective is to synthesize static function mapping the bi-level state variables to

the inputs such that with the driving safety along the route ensured, both the energy

consumption cost (i.e. fuel and electrical energy) and vehicle exhaust emissions

(𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥) are minimized. The state variables include the ones from the upper (external)

level, i.e. the accumulated driving distance of the host vehicle, 𝑠, and the ones from the

lower (internal) level, i.e. the instantaneous host vehicle velocity, 𝑣, and battery pack state

of charge (𝑆𝑂𝐶). The control inputs are the instantaneous host vehicle acceleration 𝑎, and

the split ratio between the ICE torque and the EM torque, γ. The problem is thus formulated

into a multi-objective optimization problem, as follows,

where 𝐽𝑜𝑝𝑡 is the total cost to be optimized and 𝑐(𝑥𝑘, 𝑢𝑘) is a function that maps the state

and control variables to an instantaneous cost at one timestep. Although the problem is

formulated into an infinite-horizon problem, it is meant to be finite since the system

terminates once the host vehicle arrives at the pre-defined destination. Thus, 𝐽𝑜𝑝𝑡 is

guaranteed to be a finite value.

The split ratio as one of the state variables is defined as,

min: 𝐽𝑜𝑝𝑡 =∑𝑐(𝑥𝑘, 𝑢𝑘)

∞

𝑘=1

(4. 1)

subject to: 𝑥𝑘+1 = 𝑓(𝑥𝑘 , 𝑢𝑘)

𝑥 ∈ 𝑋, 𝑢 ∈ 𝑈(𝑥) (4. 2)

γ = 𝑇𝐸𝑀/𝑇𝑟𝑒𝑞 (4. 3)

47

where 𝑇𝐸𝑀 is the torque of the EM, and 𝑇𝑟𝑒𝑞 is the demanding torque transmitted to the

ICE and EM by the gearbox.

As mentioned before, the optimization problem involves three objectives, namely, 1)

the objective of driving safety, 2) the objective of energy management and 3) the objective

of emission reduction. Therefore the cost function 𝑐(𝑥𝑘, 𝑢𝑘) can be written as,

where 𝐽𝑆𝐹(𝑥𝑘 , 𝑢𝑘), 𝐽𝐸𝑁(𝑥𝑘 , 𝑢𝑘) and 𝐽𝐸𝑋(𝑥𝑘 , 𝑢𝑘) are the instantaneous cost of driving safety,

the instantaneous cost of energy consumption and the instantaneous cost of exhaust

emission, respectively. 𝛼 and 𝛽 serve as the weighting factors to make a trade-off between

different costs.

4.2 Model predictive control-based eco-driving control

strategy

In this section, the model predictive control (MPC) based eco-driving strategy for a

connected and automated vehicle (CAV) with hybrid electric configuration, as described

in Chapter 2, is proposed. Prior to solve the problem in the context of model predictive

control (MPC) framework, the control objectives with corresponding cost functions are

introduced first, with regard to (1) the objective of driving safety that can realize the safe

driving by complying with traffic rules and keeping a safe distance with preceding vehicles

(2) the objective of energy management that aims to reduce the fuel consumption with a

balanced battery SOC level (3) the objective of emission reduction that serves to reduce

the exhaust emissions 𝐻𝐶, 𝐶𝑂 and 𝑁𝑂𝑥. Then, the objective functions are fit in the MPC

framework to derive the optimal control strategy for the vehicle.

𝑐(𝑥𝑘 , 𝑢𝑘) = 𝐽𝑆𝐹(𝑥𝑘, 𝑢𝑘) + 𝛼 ∙ 𝐽𝐸𝑁(𝑥𝑘 , 𝑢𝑘) + 𝛽 ∙ 𝐽𝐸𝑋(𝑥𝑘 , 𝑢𝑘) (4. 4)

48

4.2.1 Optimization objectives

4.2.1.1 The objective of driving safety

The goal of safe driving can be exemplified by the scenarios when the host vehicle aims to

keep a collision-free distance with other vehicles, make decisions at a TSCI, and comply

with the speed limit set by each road. In order to achieve this goal, cost penalties need to

be applied to prevent unsafe driving actions. The driving scenarios are therefore classified

into different situations to make the strategy adaptive. They are a). free driving scenario,

b). signal anticipation scenario and c). car-following scenario. The instantaneous cost of

driving safety 𝐽𝑆𝐹(𝑥𝑘 , 𝑢𝑘), as in the previous equation, is then represented as follows,

where 𝐽𝐹𝐷 , 𝐽𝑆𝐴 and 𝐽𝐶𝐹 correspond to the instantaneous cost during the free driving

scenario, signal anticipation scenario, and car-following scenario, respectively. 𝜆1, 𝜆2 and

𝜆3 are the Boolean factors ([0,1]) that turn ON/OFF for each cost. It is worth noting that at

each time step, only one out of three scenarios is triggered, hence the corresponding cost

is applied for safe driving. 𝑤1, 𝑤2 and 𝑤3 are the weighting factors to make an appropriate

trade-off between different costs.

From the V2I and V2V connections, the real-time driving data that ensure the driving

safety are ∆𝑑 , ∆𝑠 , 𝑆𝑟𝑦𝑔 , ∆𝑡𝑇𝑆 , 𝑣𝑙𝑒𝑎𝑑 and 𝑣𝑙𝑎𝑛𝑒 , as listed in the following table. The

illustration is shown in Fig. 4. 1.

𝐽𝑆𝐹(𝑥𝑘 , 𝑢𝑘) = 𝜆1 ∙ 𝑤1 ∙ 𝐽𝐹𝐷 + 𝜆2 ∙ 𝑤2 ∙ 𝐽𝑆𝐴 + 𝜆3 ∙ 𝑤3 ∙ 𝐽𝐶𝐹 (4. 5)

49

Table. 4. 1 Parameters concerning the driving safety


∆𝑑 (m) The distance to the upcoming TSCI 121.5 m

∆𝑠 (m) The distance to the closet vehicle ahead in the same lane towards

the destination

13.8m

𝑆𝑟𝑦𝑔 The traffic signal of the upcoming TSCI g

∆𝑡𝑇𝑆 (s) The remaining time for the current traffic signal phase 23s

𝑣𝑙𝑒𝑎𝑑 (m/s) The speed of the ahead vehicle in the same lane, if applies 13.8m/s

𝑣𝑙𝑎𝑛𝑒 (m/s) The speed limit for the lane where the host vehicle is running on 27.8m/s

To model the driving scenario classifier (DSC) of the studied vehicle, the general

vehicle braking distance equation is given as follows [6],

where 𝑆𝑏𝑟𝑘 denotes the braking distance in the unit of m, 𝑆𝑚𝑖𝑛 is the minimum distance to

keep in the unit of m, 𝑣(𝑘) is the host vehicle speed at time instant 𝑘 in the unit of m/s,

𝑡𝑏𝑟𝑘 is the reaction time before braking in s, 𝑎𝑏𝑟𝑘 is the braking deceleration in m/𝑠2.

∆𝑑𝐿 and ∆𝑠𝐿 are the thresholds for the TSCI distance, ∆𝑑, and inter-vehicle distance,

∆𝑠, respectively, in the unit of m. Their equation is modelled as the maximum braking

distance based on Eq. (4. 6), with parameters taken from the study [6], as follows,

when the actual distance of the host vehicle to the upcoming TSCI, ∆𝑑, is smaller than the

threshold distance, ∆𝑑𝐿, the signal anticipation scenario can be triggered by the control

system. The reactions of the host vehicle in approaching the TSCI will be anticipated.

𝑆𝑏𝑟𝑘 = 𝑆𝑚𝑖𝑛 + 𝑣(𝑘) ∙ 𝑡𝑏𝑟𝑘 +𝑣(𝑘)2

2 ∙ 𝑎𝑏𝑟𝑘 (4. 6)

∆𝑑𝐿 = ∆𝑠𝐿 = 10+ 𝑣(𝑘) + 0.0825 × 𝑣(𝑘)2 (4. 7)

50

Similarly, when the actual car-following distance, ∆𝑠, is smaller than the inter-vehicle

distance threshold ∆𝑠𝐿, the car-following scenario is triggered by the control system.

Fig. 4. 1 Illustration for driving safety parameters

The threshold used when determining if the host vehicle should pass the intersection is

determined by the following estimated passing time ∆𝑡𝑟, as follows,

where 𝑎𝑚𝑎𝑥 is the maximum allowed speed for the vehicle.

The driving scenarios are thus classified based on the above parameters in real time,

illustrated as follows, where FD, SA, and CF stand for free driving, signal anticipation and

car-following mode, respectively.

∆𝑡𝑟 =

{

∆𝑑

𝑎𝑚𝑎𝑥, 𝑣(𝑘) = 0

∆𝑑

𝑣(𝑘)𝑣(𝑘) > 0

(4. 8)

51

Fig. 4. 2 Driving scenario classifier (DSC) workflow

52

(I) Free driving

In the free driving scenario, the host vehicle is driving with speed no higher than either

the maximum speed limit imposed by the current lane or the maximum constraint speed of

the vehicle. When the vehicle accelerates to the maximum allowed speed, it then enters

into the steady speed control until the traffic conditions start to change.

The cost associated with the free driving scenario is defined as a quadratic function,

where 𝑣𝑙𝑖𝑚(𝑘) is the maximum allowed speed for the host vehicle at time step 𝑘, and it is

defined as,

which refers to the smaller one of the two, the speed limit on the lane at time step 𝑘, 𝑣𝑙𝑎𝑛𝑒,

and the maximum allowed speed of the host vehicle, 𝑣𝑚𝑎𝑥.

(II) Signal anticipation

When the host vehicle is in the signal anticipation scenario where there is a need to

decelerate depending on the conditions defined in DSC, the cost function 𝐽𝑆𝐴 is triggered

to control the subsequent reactions of the vehicle to safely reach a complete stop at the

intersection. Sample cases at the TSCI are illustrated in Fig. 4. 3,

In the cost function of the signal anticipation scenario, the minimum safe distance to

stop at the intersection, based on the braking distance modelled in Eq. (4. 6), is defined

with static minimum distance 𝑆𝑚𝑖𝑛 equaling 2 m, reaction time 𝑡𝑏𝑟𝑘 equaling 0.5 s, and

emergency deceleration equaling 8 m/𝑠2 [6],

𝐽𝐹𝐷 = (𝑣𝑙𝑖𝑚(𝑘) − 𝑣(𝑘))2 (4. 9)

𝑣𝑙𝑖𝑚(𝑘) = 𝑚𝑖𝑛 (𝑣𝑙𝑎𝑛𝑒, 𝑣𝑚𝑎𝑥) (4. 10)

∆𝑑𝑚𝑖𝑛 = 2 + 0.5 ∙ 𝑣(𝑘) + 0.0625 × 𝑣(𝑘)2 (4. 11)

53

Therefore, the cost is applied according to the distance of the vehicle to the intersection,

modelled as follows [68],

where ∆𝑑 is the actual distance of the host vehicle to the upcoming intersection and 𝑞 is

the weighting factor in the cost function. In approaching the intersection, as the ∆𝑑 gets

smaller, the penalty gets larger to help reduce the vehicle speed. If the distance of ∆𝑑

becomes less than the minimum allowed distance ∆𝑑𝑚𝑖𝑛, an infinite cost is applied to the

system.

(III) Car following

When the host vehicle enters the car-following scenario, triggered by the conditions

defined in DSC, the cost function is applied according to the inter-vehicle distance so that

a safe following distance can be achieved and any potential collision can be prevented.

The minimum and maximum following distance are derived from the Eq. (4. 7) and Eq.

(4. 11), which are relevant to the instantaneous host vehicle speed, as follows [6],

Within the boundaries of the minimum and maximum following distances, the optimal

following distances, which are bounded by the lower optimal limit ∆𝑠𝑜𝑝𝑡𝑖𝑛𝑓

and upper optimal

limit ∆𝑠𝑜𝑝𝑡𝑠𝑢𝑝

, are constructed based on the minimum and maximum following distances, the

parameters are taken from the study [68],

𝐽𝑆𝐴 = {

𝑡𝑎𝑛 (𝜋

2) , ∆𝑑 < ∆𝑑𝑚𝑖𝑛

𝑞 ∙ 𝑡𝑎𝑛 (∆𝑑 − ∆𝑑𝐿

∆𝑑𝑚𝑖𝑛 − ∆𝑑𝐿∙𝜋

2), ∆𝑑𝑚𝑖𝑛 ≤ ∆𝑑 ≤ ∆𝑑𝐿

(4. 12)

∆𝑠𝑚𝑖𝑛 = 2 + 0.5 ∙ 𝑣(𝑘) + 0.0625 × 𝑣(𝑘)2 (4. 13)

∆𝑠𝑚𝑎𝑥 = 10 + 𝑣(𝑘) + 0.0825 × 𝑣(𝑘)2 (4. 14)

54

Therefore, the cost function for the car-following scenario is derived based on the study

[6],

where 𝑓1 and 𝑓2 are the weighting factors in the function. As the vehicle enters the car-

following scenario, the penalty is applied to favor the car-following distance that falls

within the range defined by the lower optimal limit ∆𝑠𝑜𝑝𝑡𝑖𝑛𝑓

and upper optimal limit ∆𝑠𝑜𝑝𝑡𝑠𝑢𝑝

.

When the inter-vehicle distance gets less than the minimum allowed distance ∆𝑠𝑚𝑖𝑛 , an

infinite penalty will be applied.

∆𝑠𝑜𝑝𝑡𝑖𝑛𝑓

= 0.6 ∙ ∆𝑠𝑚𝑎𝑥 + 0.4 ∙ ∆𝑠𝑚𝑖𝑛 (4. 15)

∆𝑠𝑜𝑝𝑡𝑠𝑢𝑝

= 0.4 ∙ ∆𝑠𝑚𝑎𝑥 + 0.6 ∙ ∆𝑠𝑚𝑖𝑛 (4. 16)

𝐽𝐶𝐹 =

{

+∞, ∆𝑠 < ∆𝑠𝑚𝑖𝑛

𝑓1 ∙ 𝑡𝑎𝑛 (∆𝑠 − ∆𝑠𝑜𝑝𝑡

𝑖𝑛𝑓

∆𝑠𝑚𝑖𝑛 − ∆𝑠𝑜𝑝𝑡𝑖𝑛𝑓 ∙

𝜋

2) , ∆𝑠𝑚𝑖𝑛 ≤ ∆𝑠 < ∆𝑠𝑜𝑝𝑡

𝑖𝑛𝑓

|∆𝑠 −∆𝑠𝑜𝑝𝑡

𝑖𝑛𝑓+ ∆𝑠𝑜𝑝𝑡

𝑠𝑢𝑝

2| , ∆𝑠𝑜𝑝𝑡

𝑖𝑛𝑓≤ ∆𝑠 ≤ ∆𝑠𝑜𝑝𝑡

𝑠𝑢𝑝

𝑓2 ∙ (∆𝑠 − ∆𝑠𝑜𝑝𝑡𝑠𝑢𝑝)

2, ∆𝑠𝑜𝑝𝑡

𝑠𝑢𝑝 < ∆𝑠 ≤ ∆𝑠𝑚𝑎𝑥 0 ∆𝑠 > ∆𝑠𝑚𝑎𝑥

(4. 17)

55

Fig. 4. 3 Sample cases at the TSCI without leading vehicles (a) brakes are applied due

to the red signal, (b) brakes are applied because the green signal phase duration is not

sufficient for passing (c) vehicle decides to drive across the intersection in 25s

56

4.2.1.2 The objective of energy management

The cost function of the energy management objective is applied across the driving

schedule for all the driving scenarios to minimize the energy cost, i.e. the fuel energy and

the electrical energy.

The equation to calculate the instantaneous fuel cost �̇�𝑓𝑢𝑒𝑙 can be referred to as

Chapter 3.1.1.

As to the electrical energy, since the overuse of the battery will inevitably impose

intractable damage to the battery health, equivalent consumption minimization strategy

(ECMS) is adopted as an approach to bound the battery state of charge (SOC) within the

range from 0.5 to 0.75. As a result, a balance can be achieved between fuel consumption

and electrical energy consumption. The equivalent factor 𝑠𝑒 is defined as [6],

The instantaneous equivalent fuel consumption of the EM is given as [6],

where 𝑄𝑓̅̅̅̅ is the average fuel consumption rate, which is set to 240 𝑔/𝑘𝑊ℎ, �̅�𝐸𝑀 is the

average efficiency of the EM and it is assumed to be 0.88.

Therefore, the energy cost that has both the fuel consumption and the electrical energy

consumption is derived as,

𝑠𝑒 = {0.001, 𝑆𝑂𝐶 ≥ 0.75

−6 ∙ 𝑆𝑂𝐶 + 5 0.5 < 𝑆𝑂𝐶 < 0.75𝐼𝑛𝑓 𝑆𝑂𝐶 ≤ 0.5

(4. 18)

�̇�𝑒𝑚 = 𝑄𝑓̅̅̅̅ ∙ 𝑃𝑏𝑎𝑡𝑡

3.6 × 106 ∙ �̅�𝐸𝑀 (4. 19)

𝐽𝐸𝑁 = �̇�𝑓𝑢𝑒𝑙 + 𝑠𝑒 ∙ �̇�𝑒𝑚 (4. 20)

57

4.2.1.3 The objective of emission reduction

During driving, the vehicle engine mainly produces three types of exhaust pollutants

(𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥) into the air that are toxic to human health [74]. They are to be reduced by

manipulating the engine speed and torque. The cost of exhaust emissions, therefore, include

the cost of 𝐻𝐶,𝐶𝑂 and 𝑁𝑂𝑥, as follows,

The equation for the emission cost �̇�ℎ𝑐, �̇�𝑐𝑜 and �̇�𝑛𝑜𝑥 can be referred to as Chapter

3.1.1. And they are in the unit of 𝑔/𝑠.

4.2.2 Model predictive control (MPC)

Due to the simplicity and fast computation, the problem presented in Chapter 34.1 is solved

online using the bi-level nonlinear model predictive control (BL-NLMPC) strategy. Based

on the state variables at each time step, the BL-NLMPC computes a future control sequence

in the prediction horizon that minimizes the total cost in the horizon, and apply the first

element of the control sequence to the HEV model [18]. The process is repeated at each

time step with the prediction horizon moving one step forward until the system reaches the

terminal state [18]. In the control sequence, at the upper level, the controller finds the

optimal acceleration for the HEV that minimizes the total cost in the prediction horizon.

Similarly, at the lower level, the controller derives the optimal split ration between the ICE

torque and EM torque in the powertrain that minimizes the prediction horizon cost. Thus,

the optimization problem can be rewritten as follows,

𝐽𝐸𝑋 = �̇�ℎ𝑐 + �̇�𝑐𝑜 + �̇�𝑛𝑜𝑥 (4. 21)

58

where 𝑇𝑠 is the sampling time, 𝑁𝑝 denotes the receding prediction horizon. The state

variables include the upper-level states, accumulated driving distance of the ego host

vehicle, s and the instantaneous ego host vehicle velocity, and the lower-level state v,

battery pack state of charge (SOC). The control variables are the vehicle acceleration, a, at

the upper level, and the torque split ratio between ICE and EM, γ, at the lower level. To

obtain a quick control response, the control horizon 𝑁𝑐 is set to be 1, which means one set

of derived control inputs is applied across the prediction horizon. It is worth noting that the

control variables are assumed to be constant in the prediction horizon. The illustration and

workflow are described and obtained as the pseudo-code in Fig. 4. 4 and Fig. 4. 5.

𝐽𝑜𝑝𝑡 = min𝑢𝑡( ∑ ( 𝐽𝑆𝐹(𝑥𝑡 , 𝑢𝑡) + 𝛼 ∙ 𝐽𝐸𝑁(𝑥𝑡 , 𝑢𝑡) + 𝛽 ∙ 𝐽𝐸𝑋(𝑥𝑡 , 𝑢𝑡)) ∙ 𝑇𝑠

𝑘+𝑁𝑝−1

𝑡=𝑘

),

𝑘 = 0,1,2,…

(4. 22)

state variables: xk = [s, v, SOC] (4. 23)

control variables: 𝑢𝑘 = [a, γ] (4. 24)

subject to: 𝑠 > 0

0 < 𝑣 < 𝑣𝑚𝑎𝑥

𝑆𝑂𝐶𝑚𝑖𝑛 < 𝑆𝑂𝐶 < 𝑆𝑂𝐶𝑚𝑎𝑥

𝑎𝑚𝑖𝑛 < 𝑎 < 𝑎𝑚𝑎𝑥

γ𝑚𝑖𝑛 < γ < γ𝑚𝑎𝑥

(4. 25)

59

Fig. 4. 4 The framework of the proposed MPC-based eco-driving strategy

60

Fig. 4. 5 Pseudo-code of the proposed algorithm

61

4.3 Rule-based EACS

To evaluate the performance of the proposed strategy, a bi-level benchmark eco-driving

control strategy for CAHEV is designed. At the upper level, the instantaneous vehicle

speed is constructed using the intelligent driver model (IDM). At the lower level, the rule-

based electric-assist control strategy (EACS) aims to control the energy split ratio in the

electric hybrid powertrain.

4.3.1 Intelligent driver model (IDM)

As one of the most simple, comprehensive and collision-free driver models that produce

realistic profiles, the intelligent driver model (IDM) calculates the real-time acceleration

based on a set of driving conditions [75]. The acceleration is defined as follows [75],

where 𝑣 is the current vehicle speed and 𝑣0 is the target speed. 𝑎 and 𝑏 denote the

acceleration ability and deceleration capacity of the vehicle, respectively. 𝑠 is the current

distance to the obstacle, either the preceding vehicle or the TSCI. 𝑠∗(𝑣, Δ𝑣) is the desired

distance which is defined in Eq. (4. 28). 𝑇 is the time gap that is set to 1.0 s. Δ𝑣 is defined

as the speed difference between the host vehicle current speed 𝑣 and the obstacle speed

𝑣1. If the obstacle ahead is a preceding vehicle, 𝑣1 > 0, or 𝑣1 = 0 for the TSCI. 𝑠0 is the

minimum allowed gap between the host vehicle and the obstacle. The parameters are

selected according to [75] in combination with an estimation of real-world vehicles, which

are listed in the following table.

�̇� = 𝑎 [1 − (𝑣

𝑣0)𝛿

− (𝑠∗(𝑣, Δ𝑣)

𝑠)2

] (4. 26)

Δ𝑣 = 𝑣 − 𝑣1 (4. 27)

𝑠∗(𝑣, Δ𝑣) = 𝑠0 +max (0, 𝑣𝑇 +𝑣Δ𝑣

2√𝑎𝑏) (4. 28)

62

Table. 4. 2 Parameters for IDM

Intelligent Driver Model

Parameters Symbols Values

Time gap (s) 𝑇 1.0

Minimum gap (m) 𝑠0 2

Acceleration exponent (-) 𝛿 4

Acceleration (m/s2) 𝑎 2

Comfortable deceleration (m/s2) 𝑏 2

4.3.2 Electric-assist control strategy (EACS)

The electric-assist control strategy (EACS) is one of the most commonly used and practical

rule-based strategies for a parallel HEV [76], in which the ICE is used as the main power

source while the battery provides assistant power. The EM runs either when the torque

demand is too high, or when the ICE is operating inefficiently. As a result, the total fuel

economy can be improved over the trip. A visualization of the rules in EACS is shown in

Fig. 4. 6.

The control strategy determines the power split between ICE and EM in the powertrain,

with the driving conditions taken into consideration. The application of EACS can help

achieve a better fuel economy and fewer exhaust emissions [76]. The variables that

determine the rule-based policies are defined in Table. 4. 3 [45]. The basic rules in EACS

can be described as follows,

1. If SOC is higher than the upper bound SOCH, then ICE turns off, EM turns on

2. If SOC is lower than the upper bound SOCH and higher than the lower bound SOCL,

63

(1) If the vehicle speed is lower than the vehicle low-speed threshold VL or

if the demand torque Treq is lower than the ICE off torque curve, then

ICE turns off, EM turns on

(2) Otherwise, if the demand torque Treq is lower than the maximum ICE

torque TICEmax, then ICE turns on, EM turns off

(3) Otherwise, if the demand torque Treq is higher than the maximum ICE

torque TICEmax , then ICE turns on, also EM turns on to assist ICE in

supplying the power by TEM = Treq − tdis ∙ TICEmax where tdis is the

discharge factor

3. If SOC is lower than the lower bound SOCL

(1) If the demand torque Treq is lower than the ICE minimum torque curve,

then ICE turns to drive the vehicle while charging the battery by

redundant torque defined as tmin ∙ TICEmax − Treq where tmin denotes the

ICE minimum torque threshold factor, EM turns on to switch to

generator mode

(2) If the demand torque Treq is higher than the ICE minimum torque curve,

then ICE turns to drive the vehicle while charging the battery by

redundant torque defined as tch ∙ TICEmax where tch denotes the ICE

charging factor, EM turns on to switch to generator mode

(3) If the demand torque Treq is higher than the maximum ICE torque TICEmax,

there is not enough power to drive the vehicle

4. If the vehicle is completely stopped, then ICE turns off, EM turns on

64

5. If the vehicle is braking, then ICE turns off, EM turns on to switch to generator

mode to charge the battery

Fig. 4. 6 EACS rule-based illustrations

65

Table. 4. 3 Parameters for rule-based EACS

Parameter Description

𝑆𝑂𝐶𝐿 The lowest SOC allowed

𝑆𝑂𝐶𝐻 The highest SOC allowed

𝑡𝑜𝑓𝑓 The threshold factor for the ICE to shut off during 𝑆𝑂𝐶𝐿 < 𝑆𝑂𝐶 < 𝑆𝑂𝐶𝐻

𝑡𝑚𝑖𝑛 The threshold factor for the ICE to charge the EM

during 𝑆𝑂𝐶 < 𝑆𝑂𝐶𝐿

𝑡𝑐ℎ The charging factor for the ICE during 𝑆𝑂𝐶𝐿 < 𝑆𝑂𝐶 <𝑆𝑂𝐶𝐻

𝑡𝑑𝑖𝑠 The discharging factor for ICE during 𝑆𝑂𝐶𝐿 < 𝑆𝑂𝐶 <𝑆𝑂𝐶

𝑉𝐿 The lowest vehicle speed threshold

The algorithm is described in the following pseudo-code,

66

Fig. 4. 7 Pseudo code of rule-based EACS

67

The policy is illustrated in Fig. 4. 8.

Fig. 4. 8 Workflow of the EACS

4.4 Summary

In this chapter, the eco-driving problem has been formulated first. The goal is to achieve a

fuel-efficient and low-emission trip on the basis of driving safety.

To achieve this goal, a bi-level model predictive control (MPC) based control strategy

is designed for a connected and automated hybrid electric vehicle (CAHEV). In this

framework, the problem is first modelled as multi-objective optimization problem with

68

regard to (1) the objective of driving safety that can realize the safe driving by complying

with traffic rules and keeping a safe distance with preceding vehicles (2) the objective of

energy management that aims to reduce the fuel consumption with a balanced battery SOC

level (3) the objective of emission reduction that serves to reduce the exhaust emissions

𝐻𝐶, 𝐶𝑂 and 𝑁𝑂𝑥. Next, a cost function is assigned to each objective. At the upper level of

the control strategy, based on the driving variables via V2I and V2V communications, a

driving scenario classifier is built up to recognize the instantaneous driving case so as to

assign the corresponding cost in the optimization function. Then, the optimal vehicle

acceleration that determines the velocity at the next time step is computed. At the lower

level, given the torque demand on the wheels, the control system derives the optimal torque

split ratio between ICE and EM to enable the powertrain working in a high-efficiency state.

As a result, the overall fuel consumption and the engine-out emissions can be minimized,

with the driving safety always ensured over the trip.

In order to evaluate the performance of the proposed eco-driving control strategy, the

rule-based electric-assist control strategy (EACS), as one of the widely applied approaches

for HEV energy management, in combination with the intelligent driver model (IDM) is

synthesized as a comparison. In this benchmark strategy, The IDM model decides the

realistic driving velocity at each time instant over the trip while the rule-based EACS

computes the real-time torque split ratio in the powertrain based on the torque demand.

69

Chapter 5. Results and Discussion

5.1 Experiment setup

The simulation is running in MATLAB and SUMO, with the vehicle parameters obtained

from ADVISOR. The background map for the studied vehicle with realistic traffic demand

is from the TAPAS-Cologne project [77].

5.1.1 ADVISOR

ADVISOR is the short term for “Advanced Vehicle Simulator.” It is a vehicle powertrain

simulator designed by National Renewable Energy Laboratory (NREL) in 1994 [78].

ADVISOR is based on empirical drivetrain components to evaluate vehicle performance

under given driving cycles, such as fuel economy and exhaust emissions [78]. The

drivetrain types in ADVISOR ranges from the ones for conventional internal combustion

engine vehicles, hybrid electric vehicles, plug-in hybrid electric vehicles to pure electric

vehicles. In this research, the studied parallel hybrid electric vehicle data is obtained from

ADVISOR, including the data of internal combustion engine, electric motor, battery pack,

transmission shift ratio, and vehicle physical parameters. The sample ADVISOR user

interface is presented in Fig. 5. 1,

70

Fig. 5. 1 Sample ADVISOR interface for customizing vehicle parameters, a vehicle

with a parallel hybrid powertrain is configured

5.1.2 SUMO

SUMO stands for “Simulation of Urban MObility,” and it is an open-source platform that

is able to simulate realistic microscopic and macroscopic urban traffic flow mobility [79].

In this environment, an individual vehicle is considered as an agent that is running in the

simulated map. The attributes of the agent can be customized to mimic the real-world

vehicle, such as the vehicle type, the vehicle length, the speed limit, the acceleration limit,

the frontal surface area and the like. On the other hand, the urban simulation map that can

be either modelled using the built-in tool, or imported from the real world map involves

71

most realistic traffic elements, from the highway and urban roads, bus stops, induction loop

to the road speed limit, customizable traffic lights program, and right-of-way rules, etc.

SUMO was first initiated in 2001 by a ground of researchers associated with the

German Aerospace Center (DLR) and some universities. Within the past ten years’

development, SUMO has become a comprehensive suite for traffic simulation. Some of

the main tools from SUMO are listed:

(1) NETEDIT: a tool that is capable of reading and importing the real-world maps

from different sources into SUMO

(2) OD2TRIPS: that can map the traffic demand via the format of Origin/Destination

(O/D) Matrices into individual vehicle trips on the map.

(3) TraCI: short term for “Traffic Control Interface”, which allows the online TCP-

based communication with the simulated objects, such as running vehicles, traffic

lights, and the online manipulation via commands

(4) SUMO-GUI: the graphical user interface to display the online traffic simulation

(Fig. 5. 2 shows a sample simulation on SUMO-GUI)

In this research, SUMO is used as the main platform to simulate and verify the bi-level

eco-driving strategy for a connected and automated hybrid electric vehicle.

72

Fig. 5. 2 Sample SUMO-GUI interface, zoomed-out view of the Cologne, Germany

urban transportation network

Fig. 5. 3 A zoomed-in view of the Cologne, Germany urban transportation network

[77]

73

5.1.3 Traffic simulation map

The simulation is running in MATLAB and SUMO on a PC with a 1.8 GHz Intel Core i5-

8250U CPU and 16 GB of internal memory. The host connected and automated vehicle

(CAV) is modelled as an HEV with a single-shaft pre-transmission hybrid electric

powertrain, as depicted in Fig. 3. 1. The vehicle specifications are obtained from

ADVISOR. Unlike the host CAV, the background vehicles in traffic are assumed to be of

the same type - conventional passenger vehicles that are driven by a human. Their dynamic

actions, including the instantaneous speed and acceleration, are determined by the car-

changing model. The specifications are given in Table. 5. 1.

The simulation uses an urban map of Cologne, a typical medium-sized city in Germany,

with a realistic traffic demand dataset from the TAPAS-Cologne project by the Institute of

Transportation Systems at the German Aerospace Center (ITS-DLR) [77]. The road

network covers the area of Cologne approximately 400 𝑘𝑚2 formed by 71,368 road

segments. The realistic 24-hour traffic dataset, comprising of more than 700,000 individual

trips, faithfully models the daily traffic activities of the drivers from Cologne [80]. Table.

5. 2 lists the attributes of the map. Fig. 5. 4 presents the city map of Cologne.

The host CAV starts from a complete stop (i.e. speed begins with 0 m/s and acceleration

begins with 0 m/s2) with battery state of charge at the level of 60%. As a result, the initial

state variables can be represented as [0,0,0.6]. At each time instant, the longitudinal speed

and acceleration of the host CAV are controlled by the proposed MPC-based strategy. The

lateral dynamics is controlled by a simple lane-changing model, which triggers a lane-

changing signal to the lane that leads the vehicle to the next road on the trip every time the

vehicle enters a new road, and then keeps the vehicle stay in the lane until the next

intersection. To evaluate the energy performance of the proposed eco-driving strategy, the

host CAV is also controlled by the IDM strategy on the same trip, and the EACS rule-based

strategy is applied on the vehicle powertrain to determine the instantaneous split ratio

between the mechanical power and electrical power at each time instant.

74

The parameters for the constraint optimization problem, the proposed strategy, and the

rule-based strategy are listed in Table. 5. 3

Fig. 5. 4 Urban network map of Cologne, Germany [77]

75

Table. 5. 1 Background vehicle specifications

Components Specification

Architecture

Vehicle Length

Acceleration ability (m/𝒔𝟐)

Deceleration ability (m/𝒔𝟐)

Conventional passenger car

4.3 m

2.6 m/𝑠2

-4.5 m/𝑠2

Lane-changing Model LC2013 [81]

Car-following Model Krauss [82]

Table. 5. 2 Specifications of the simulation map

City The metropolitan area of

Cologne, Germany

Area (𝒌𝒎𝟐) approx. 400 k𝑚2

No. of Roads 71,368

No. of Intersections 10,153

No. of individual Trips Over 700,000

Duration 24 hours

76

Table. 5. 3 Parameters for simulation

HEV constraints


Maximum speed (km/h) 𝑣𝑚𝑎𝑥 162

Deceleration ability (m/s2) 𝑎𝑚𝑖𝑛 -5

Acceleration ability (m/s2) 𝑎𝑚𝑎𝑥 5

Minimum SOC (%) 𝑆𝑂𝐶𝑚𝑖𝑛 0

Maximum SOC (%) 𝑆𝑂𝐶𝑚𝑎𝑥 100

Split ratio lower bound (-) γ𝑚𝑖𝑛 -10

Split ratio upper bound (-) γ𝑚𝑎𝑥 10

MPC-Based Eco-Driving Strategy


Sampling time (s) 𝑇𝑠 1

Prediction horizon (s) 𝑁𝑝 5

Control horizon (s) 𝑁𝑐 1

The weighting factor for energy cost (-) 𝛼 2.5

The weighting factor for emission cost (-) 𝛽 8

The driving safety weight factor I (-) 𝑤1 1

The driving safety weight factor II (-) 𝑤2 1

The driving safety weight factor III (-) 𝑤3 1

The driving safety weight factor IV (-) 𝑞 0.01

The driving safety weight factor V (-) 𝑓1 0.2

The driving safety weight factor VI (-) 𝑓2 2

EACS Rule-Based Strategy


The lowest SOC allowed (-) 𝑆𝑂𝐶𝐿 0.5

The highest SOC allowed (-) 𝑆𝑂𝐶𝐻 0.7

The threshold for ICE shut off (-) 𝑡𝑜𝑓𝑓 0.2

The charging threshold for low SOC (-) 𝑡𝑚𝑖𝑛 0.5

The charging torque for normal SOC (-) 𝑡𝑐ℎ 50

The discharging factor (-) 𝑡𝑑𝑖𝑠 0.2

The threshold for vehicle speed (m/s) 𝑉𝐿 4

77

5.2 Simulation results on a highway-urban trip

5.2.1 Trip Setup

A 10 km-trip is chosen to perform the simulation, as shown in Fig. 5. 5. The trip begins

from a start point located at the suburban area (Point A in Fig. 5. 5) to the destination in

the urban area (Point B in Fig. 5. 5) in Cologne, Germany around 7:30 in the morning,

which mimics a daily home-work commute. The roads that constitute the route are chosen

based on the shortest distance between the start and the endpoint. Point P that lies in the

middle of the trip is the border point between the suburban routes which are mainly made

up of highspeed highway roads (between Point A to Point P), and the urban routes which

are mainly made up of urban roads (between Point P to Point A). Details of the trip are

provided in the following table.

78

Fig. 5. 5 The simulated highway-urban trip from point A to point B that passes point

P [77]

Table. 5. 4 Specifications of the highway-urban trip

Highway-Urban Trip

Total Length (𝒌𝒎) 10.407 𝑘𝑚

No. of Roads 49

Highway / A – P (𝒌𝐦) 6.862 𝑘m

Urban / B – P (𝒌𝐦) 3.545 𝑘m

79

The speed profiles produced by the proposed eco-driving strategy and the IDM + EACS

are presented in Fig. 5. 6 and Fig. 5. 7, respectively.

Fig. 5. 6 Speed profile by MPC-based eco-driving strategy in on an highway-urban

trip

Fig. 5. 7 Speed profile by the rule-based EACS strategy on an highway-urban trip

Highway-urban trip (7:30am – 7:51am)

Free driving

Signal anticipation

Car-following

Speed profile

Speed profileHighway-urban trip (7:30am – 7:51am)

80

In Fig. 5. 8, it is clearly shown that the highway driving (from point A to point P over

the trip in Fig. 5. 5) corresponds to the speed profile from 0 s to around 300th s, and the rest

of the speed profile mainly consists of urban driving with congested traffic conditions.

According to the driving scenario classifier, the complete trip can be broken down into

three scenarios triggered by the real-time driving parameters listed in Table. 4. 1, they are

1) free driving scenario 2) Signal anticipation scenario and 3) car-following scenario,

illustrated in Fig. 5. 6. The speed profile by running IDM for EACS, as shown in Fig. 5. 7,

on the other hand, results in a collision-free safe driving along the same trip.

5.2.2 Driving Safety

5.2.2.1 Free driving

With the cost for free driving described in Chapter 4.2.1.1 applied, the resultant vehicle

driving in Fig. 5. 8 demonstrates that, first, before reaching the maximum speed limit, the

control system is able to make vehicle speed gradually converge to the target speed when

there is no vehicle, secondly, by adding extra penalties on the maximum allowed speed,

the control system can always ensure that the instantaneous vehicle speed strictly complies

with the maximum allowed speed set by each road where the vehicle is running on.

Moreover, in the zoomed-in region in Fig. 5. 8, J, K, and L represent three connecting

roads on the trip with the maximum allowed speed as 50 km/h, 30 km/h and 50 km/h,

respectively. Combined with Fig. 5. 6, it can be observed that the speed fluctuation on the

road J was caused by the car-following actions. On entering the road K, the vehicle is

switched to the free driving mode by the driving scenario classifier, and its speed gradually

reaches the speed limit and keeps constant until, at the end of this road, it enters the next

road L with a higher speed limit. Then the vehicle accelerates to gradually adapt its speed

to the road speed limit of 50 km/h on the road L and stays constant in the free driving mode.

81

The above proves the proposed control strategy can execute an adaptive and safe control

of the vehicle speed throughout the entire driving trip.

5.2.2.2 Signal anticipation

Over the entire trip, the vehicle has driven through seven TSCIs in total, as illustrated in

the red shaded regions in Fig. 5. 9. The control strategy for signal anticipation is designed

so as to stop the vehicle at the yellow and red lights, or the green lights with insufficient

remaining phase duration to pass, and let the vehicle pass the green signal intersection

given enough remaining phase time, as modelled in Chapter 4.2.1.1.

The zoomed-in region of Fig. 5. 9 shows the vehicle speed profile at a TSCI where the

signal has gone through the phases from green, yellow, red and green for the next round. It

can be clearly observed that the signal is green before the driving scenario classifier turns

on the signal anticipation mode for the vehicle. At the moment when the vehicle starts to

react to the traffic lights, though it is still green which permits the passing, the control

system sends the braking command based on the result that the estimated passing time ∆𝑡𝑟

(in Eq. (4. 8)) for the vehicle to pass the intersection turns out to be more than the remaining

green signal duration; in other words, the signal will turn red before the vehicle is able to

drive through the TSCI. As a result, the vehicle braking command causes the deceleration

in green phase #1 and yellow signal phase #2 in the figure. Then the vehicle remains

completely stopped during the red signal phase #3. When the signal turns to green again,

as denoted as phase #4, the control signal compares the estimated passing time ∆𝑡𝑟 and the

remaining phase time, and sends out the pass command, which triggers the vehicle start-

up until the vehicle leaves the intersection. In this way, safe automated driving that

conforms to the traffic rules are guaranteed, especially, the use of estimated passing time

helps to anticipate the travel time and avoid the unnecessary emergency braking at the

intersection when the remaining signal phase time is insufficient for the vehicle to pass.

82

5.2.2.3 Car-following

The control strategy for the car-following scenario, as described in Eq. (4. 17), works

toward keeping the following distance between the host and preceding vehicle at a safe

range bounded by the minimum and maximum allowed following distances, which are

changing with instantaneous vehicle speed over time, as described in Eq. (4. 13) and Eq.

(4. 14). Fig. 5. 10 depicts the car-following distances with shaded regions representing

different driving scenarios over the trip. The car-following distances thereby are only

displayed in the yellow shaded regions as car-following scenarios. It’s worth noting that

each yellow shaded segment denotes a complete car-following action with a single

preceding vehicle. Due to the dynamic and mixed traffic conditions in the realistic

simulation map (as can be seen in Fig. 5. 6), the car-following scenario of the host vehicle

can be interrupted by different real-time cases, such as the preceding vehicle left the lane

or the road, or the host vehicle turns to another road. The host vehicle driving mode is then

switched accordingly until another preceding vehicle is detected in the range threshold ∆𝑠𝐿

which triggers a next car-following mode.

It can be clearly seen from Fig. 5. 10 that the following distance of the host vehicle

never violates the upper and lower bounds. As a result, the safe car-following requirement

is satisfied. In the zoomed-in region, the region P, Q and R are the car-following scenarios,

the free-driving scenario, and the car-following scenario, respectively. In the car-following

scenario P, the vehicle keeps a safe distance with the preceding vehicle, then completes the

following action as the preceding vehicle changes to another lane. As no preceding vehicle

is found in the detection range threshold ∆𝑠𝐿, the vehicle is switched to the free driving

mode in scenario Q. Soon, as a new preceding vehicle is found in scenario R. The host

vehicle starts to keep a safe distance with the preceding vehicle in R. The upper and lower

bound distance bounds vary as the host vehicle speed changes throughout the trip.

With the car-following control strategy applied, the host vehicle is able to keep a safe

distance from the preceding vehicle and initiates adaptive car-following actions across the

whole trip.

83


throughout an highway-urban trip

Fig. 5. 9 Signal anticipation by MPC-based eco-driving strategy on an highway-urban

trip

Highway-urban trip (7:30am – 7:51am)

Speed profile

Road speed limit

J K L

300

Highway-urban trip (7:30am -7:51am)

Signal anticipation

Speed profile

#1 #2 #3 #4

84

Fig. 5. 10 Car-following scenarios by MPC-based eco-driving strategy on an

highway-urban trip

Car-following distance (7:30am – 7:51am)

Upper distance limit

Car following distance

Lower distance limit

Free driving

Signal anticipation

Car-following

P Q R

85

Fig. 5. 11 The MPC-based eco-driving strategy results including (a) Brake-specific

fuel consumption (BSFC), (b) Efficiency of the electric motor (EM) and (c) SOC of the

battery pack on an highway-urban trip

86



battery pack on an highway-urban trip

87


strategy and the EACS rule-based strategy on an highway-urban trip

Fig. 5. 14 The exhaust emissions of HC, CO, NOx produced by the MPC-based eco-driving

strategy and the EACS rule-based strategy on an highway-urban trip

Exhaust Emissions

Em

issi

on

s (g

/km

)

0.00

5.00

10.00

15.00

20.00

HC CO NOx

1.227

4.063

0.3991.763

14.663

0.535

EACS MPC-based eco-driving

88

5.2.3 Energy consumption and emission reduction

The energy performances of the MPC-based strategy and the EACS rule-based strategy

can be referred to in Fig. 5. 11 and Fig. 5. 12. In terms of engine efficiency, as shown in

Fig. 5. 11 (a) and Fig. 5. 12 (a), the dashed line represents the engine optimal engine

working line. Compared with the EACS rule-based strategy, the engine operation points

produced by MPC-based strategy tends to aggregate in the high-efficiency region,

illustrated as the red circled region on the left. While for the operations of the EACS rule-

based strategy, due to the fact that a rigid set of rules rather than an accurate model-based

optimizer is used, the engine working points tend to be more evenly distributed over the

entire operation regions without a propensity for high-efficiency region. For this reason,

the engine operation of the EACS rule-based strategy over the entire simulation duration

results in a suboptimal efficiency. As a complement of Fig. 5. 11 (a) and Fig. 5. 12 (a),

Fig. 5. 13 shows the distribution of engine working points in regions with different

efficiency, where column 240~270 (g/kWh), column 270~500 (g/kWh) and column

500~900 (g/kWh) denote the high, middle, low BSFC map efficiency regions, respectively.

It can be seen that for the MPC-based strategy, the distribution of working points in the

high-efficiency region (240~270 (g/kWh)) is 47.49%, which is significantly higher than

those from EACS strategy (16.71%). Besides, in the middle (270~500 (g/kWh)) and low

(500~900 (g/kWh)) efficiency regions, the MPC-based strategy has working points that

account for 52.21% and 0%, which are both fewer than those of the EACS strategy, as

83.01% and 0.28%, respectively. It means that the engine of the proposed MPC-based

strategy works much efficiently than the EACS strategy over the same trip, resulting in a

higher aggregation of high-efficiency working points in the figure.

In terms of the electric motor efficiency, as seen in Fig. 5. 11 (b) and Fig. 5. 12 (b) for

the MPC-based strategy and the EACS rule-based strategy, respectively. It is clearly

observed that the motors controlled by both methods take good use of regenerative braking

kinetics to charge the battery pack, as indicated by the region corresponding to the negative

motor torque. However, the motor working of the MPC-based strategy performs better than

89

the EACS rule-based strategy in the yellow circled region corresponding to the high-

efficiency region, which demonstrates that the motor under the control of the MPC-based

strategy is working more efficiently to assist the engine over the entire trip, thus saves more

electrical energy.

In comparison to the SOC of the EACS rule-based strategy shown in Fig. 5. 12 (c), the

SOC of the MPC-based strategy based on the current set of parameters (Fig. 5. 11 (c))

shows that a more energy sustaining policy is adopted to propel the vehicle. Therefore, a

healthy battery SOC is balanced. Despite that, the terminal SOC of the MPC-based strategy

shows that the battery still gets charged properly by good use of regenerative braking

kinetic energy along with the redundant energy from the engine over the trip.

The numeric results from the proposed MPC-based strategy and the EACS rule-based

strategy are presented in Table. 5. 5. It is obvious that the overall fuel consumption from

the proposed strategy is reduced by 26.80% (SOC corrected) with respect to the one of the

EACS rule-based strategy, which is a great improvement. Similar to the terminal SOC as

54.05% of the EACS rule-based strategy, the 57.74% terminal SOC of the MPC-based

strategy shows that the proposed strategy is more likely to sustain the electrical energy to

a healthy level when powering the vehicle, at the same time, the SOC level is much higher

than the minimum allowed SOC level, 0.5, where a buffering room is left for battery use

over the trip.

The vehicle emissions from the MPC-based strategy are also reduced by 27.69%,

73.91% and 47.59% for exhaust emissions ℎ𝑐, 𝑐𝑜 and 𝑛𝑜𝑥, respectively, with respect to

the results from the EACS rule-based strategy. The exceptional high saving in the 72.30%

𝑐𝑜 consumption produced by the MPC-based strategy comes from the fact that in

comparison to the engine working points from the EACS rule-based strategy (Fig. 5. 12

(a)) much less engine operation points under the control of MPC-based strategy Fig. 5. 11

(a)) tend to gather in the green circled region which corresponds to the high 𝑐𝑜 cost region.

The exhaust emission chart can be found in Fig. 5. 14.

90

The results over a day-to-day work-home commute trip demonstrate that the proposed

MPC-based eco-driving strategy is capable of keeping the vehicle driving safely in

highway and urban roads, while greatly improving the energy economy as well as reducing

the vehicle emissions when the IDM + rule-based EACS strategy is used as the benchmark.


rule-based strategy (SOC corrected) on a highway-urban trip

Strategies EACS MPC-based Improvement

Fuel consumption (g) 796.94 583.35 -26.80%

Terminal SOC (%) 54.05 57.74 ---

HC 6.61 4.78 -27.69%

CO 195.13 50.90 -73.91%

NOx 13.68 7.17 -47.59%

5.3 Simulation results on a suburban-urban trip

5.3.1 Trip Setup

In order to verify the result of the proposed strategy, another 10 km-trip is selected with

road conditions consisting of mainly suburban and urban road types, which are different

from the highway-urban trip. The trip details are presented in the following figure and table.

91

Fig. 5. 15 The simulated suburban-urban trip from point A to point B [77]

Table. 5. 6 Specifications of the suburban-urban trip

Highway-Urban Trip

Total Length (𝒌𝒎) 10.109 𝑘𝑚

No. of Roads 97

Suburban (𝒌𝐦) 2.347 𝑘m

Urban (𝒌𝐦) 7.762 𝑘m

92

The parameters remain the same as the previous trip, which can be seen in Table. 5. 3.

The speed profiles produced by the proposed eco-driving strategy and the IDM + EACS

are presented in Fig. 5. 15 and Fig. 5. 16, respectively.

Fig. 5. 16 Speed profile by MPC-based eco-driving strategy on a suburban-urban trip

Fig. 5. 17 Speed profile by the rule-based EACS strategy on a suburban-urban trip

Suburban-urban trip (7:30am – 7:54am)

Free driving

Signal anticipation

Car-following

Speed profile

Speed profile


93

It is obvious in Fig. 5. 16 and Fig. 5. 18 that the trip is mixed with both the suburban

road conditions (which are labelled as road segment #1, #2, #3 and #4 in Fig. 5. 18) and

urban road conditions. The coloured regions in Fig. 5. 16 represent different driving

scenarios switched by the on-board driving scenario classifier (DSC) in the control strategy.

On the other hand, the speed profile produced by IDM results in a collision-free driving

cycle over the trip.

5.3.2 Driving Safety

5.3.2.1 Free driving

With the cost of free driving applied in this strip, the speed profile in Fig. 5. 18 shows that

the control system is able to prevent the instantaneous vehicle speed from violating the

speed limit set by the roads where the vehicle is running on.

In addition, as displayed in the zoomed-in region in Fig. 5. 18, J, K, and L are three

connecting roads on the trip with speed limit as 70 km/h, 50 km/h and 70 km/h, respectively.

It can be seen that as the vehicle approaches the speed limit on the road J, it turns to the

road K and gradually adapts the speed to the speed limit of 50 km/h. When the vehicle turns

to the next road L, it accelerates to the speed limit of 70 km/h until the speed was stopped

by a TSCI. The vehicle continues to drive when the signal turns to green. The above shows

the effectiveness of the proposed control strategy in complying with the speed limits during

the free driving mode.

5.3.2.2 Signal anticipation

There exist more TSCIs over this trip when compared to the highway-urban trip, as shown

in Fig. 5. 19. The control strategy that aims to determine whether the pass/brake command

94

based on the current signal phase and phase remaining time is applied. It can be seen in the

zoomed-region in Fig. 5. 19 that, after the control strategy receives the traffic signal-related

data at entering the anticipating region, the brake command is executed since there is not

sufficient time for the vehicle to pass the intersection even if the signal is green at the

moment. The vehicle speed brakes to a complete stop in green phase #1 and remains static

in yellow signal phase #2 and red signal phase #3 until the signal turns to green again with

sufficient passing time for the vehicle. Then the vehicle starts up to pass the TSCI. The

results from the signal anticipation principles show that safe vehicular actions can be

expected at a TSCI with the help of data received via V2I communication.

5.3.2.3 Car-following

In the car-following scenario, the main goal of the vehicle is to keep a safe distance with

the preceding vehicle during driving. The distance varies with instantaneous vehicle speed

because a corresponding safe braking distance has to be set at different vehicle speeds. In

Fig. 5. 20, the yellow shaded regions are the car-following cases recorded throughout this

trip. Due to the mixture of driving conditions, the car-following actions could be

interrupted by different cases. Over the trip, it’s clearly shown that the inter-vehicle

distances are kept in the upper and lower bounds of the allowed distance.

It can be observed that in the zoomed-in region in Fig. 5. 20, the vehicle first completes

a safe car-following distance to a preceding vehicle, as denoted as case P, until a TSCI that

stops the vehicle from following, in case Q. As the signal permits passing, the vehicle

enters the free driving mode and catches a preceding vehicle and switches to the next car-

following scenario, as shown in case R and S, respectively.

Therefore, the car-following mode controlled by the strategy enables the vehicle to

have a collision-free and accident-free driving trip.

95


throughout a suburban-urban trip

Fig. 5. 19 Signal anticipation by MPC-based eco-driving strategy on a suburban-

urban trip


Speed profile

Road speed limit

J K L

#1 #2 #3 #4


Signal anticipation

Speed profile

#1 #2 #3 #4

96

Fig. 5. 20 Car-following scenarios by MPC-based eco-driving strategy on a

suburban-urban trip

Car-following distance (7:30am – 7:54am)

Upper distance limit

Car following distance

Lower distance limit

Free driving

Signal anticipation

Car-following

P Q R S

97

Fig. 5. 21 The MPC-based eco-driving strategy results including (a) Brake-specific

fuel consumption (BSFC), (b) Efficiency of the electric motor (EM) and (c) SOC of the

battery pack on a suburban-urban trip

(a) (b)

(c)

0.5519

98



battery pack on a suburban-urban trip

(a) (b)

(c)

0.5438

99


strategy and the EACS rule-based strategy on a suburban-urban trip

Fig. 5. 24 The exhaust emissions of HC, CO, and NOx produced by the MPC-based

eco-driving strategy and the EACS rule-based strategy on a suburban-urban trip

Exhaust Emissions

Em

issi

on

s (g

/km

)

0.00

5.00

10.00

15.00

20.00

HC CO NOx

1.02

3.482

0.3431.472

12.766

0.497

EACS MPC-based eco-driving

100

5.3.3 Energy consumption and emission reduction

The energy performances of the MPC-based strategy and the EACS rule-based strategy

can be seen in Fig. 5. 21 and Fig. 5. 22. It can be clearly seen that, with the proposed

strategy applied, the engine can work in high efficiency over the suburban-urban trip, as

depicted by the aggregation of engine working points found in the high-efficiency region

270 (g/kWh). Whereas for the results by the rule-based EACS strategy, the engine is

working in an average state with operation points evenly distributed over the BSFC map.

This can be explained by the nature of both strategies as the rule-based are designed by

heuristic rules from experience or human expertise that are unable to be adapted on a case-

by-case trip, while MPC-based control strategy aims to derive the optimal control results

based on the information received in real-time, thus is more adaptive and yields better

performance.

As a complement of Fig. 5. 21 (a) and Fig. 5. 22 (a), Fig. 5. 23 shows the distribution of

engine working points in regions with different efficiency, where column 240~270

(g/kWh), column 270~500 (g/kWh) and column 500~900 (g/kWh) correspond to the high,

middle, low BSFC map efficiency regions, respectively. It is shown in the figure that over

the suburban-urban trip, the engine efficiency from the proposed control strategy is

significantly higher than that from the EACS as working points in high-efficiency region

(240~270 (g/kWh)) accounts for 85.07% of the entire working points, where they are 10.89%

in EACS strategy.

Table. 5. 7 displays the numeric results. In comparison to the fuel consumption from

the EACS, the fuel consumption from the proposed approach is greatly reduced by 38.15%.

The SOC from both strategies are close in number, showing an energy-sustaining ability

in the proposed strategy and the rule-based strategy.

The vehicle emissions from the MPC-based strategy are also decreased by 31.08%,

72.72% and 30.73% for exhaust emissions ℎ𝑐, 𝑐𝑜 and 𝑛𝑜𝑥, respectively, when compared

101

with the results from the rule-based strategy. The visualization of emission reduction is

shown in Fig. 5. 24.

In sum, the performance over a day-to-day suburban-urban trip show that, when

compared to the widely applied rule-based strategy along with IDM for realistic speed

profile generation, the vehicle controlled by the proposed eco-driving approach is able to

achieve much better fuel economy and less engine-off emissions, while the safe driving is

enabled over the entire trip.


rule-based strategy (SOC corrected) on a suburban trip

Strategies EACS MPC-based Improvement

Fuel consumption (g) 671.21 415.14 -38.15%

Terminal SOC (%) 54.38 55.19 ---

HC 5.02 3.46 -31.08%

CO 128.94 35.17 -72.72%

NOx 14.87 10.30 -30.73%

5.4 Summary

The chapter examines the performance of the proposed strategy. First, a realistic urban map

is set up to simulate a 24hr traffic condition as the background for the test. In order to

evaluate the results of the proposed strategy, EACS, as a widely used rule-based strategy

for HEV, and IDM, which is an intelligent driver model to generate realistic collision-free

speed profile, are used as the benchmark. Next, a simulation over a selected highway-urban

trip is carried out on both strategies. The results show that the proposed MPC-based

strategy is able to ensure a safe driving distance to the preceding vehicle, legal speeds kept

102

on the road and anticipating actions on a TSCI. In addition, when compared with the

benchmark strategy, the fuel consumption from the proposed strategy is greatly reduced by

26.80%, and much fewer exhaust emissions are produced, by 27.69%, 73.91% and 47.59%

for ℎ𝑐, 𝑐𝑜 and 𝑛𝑜𝑥, respectively, which demonstrates a great improvement. Moreover, a

simulation on a suburban-urban trip is carried out to verify the approach. When compared

to the EACS, The results from the MPC-based eco-driving achieve an even bigger

improvement in fuel economy, by 38%, and much fewer emissions are generated, by

31.08%, 72.72% and 30.73% for ℎ𝑐, 𝑐𝑜 and 𝑛𝑜𝑥, respectively, which show the robustness

and adaptability of the proposed strategy. The proposed eco-driving strategy in this study

is expected to be implemented to a vehicle onboard control device of a real-world CAHEV

in highway and urban driving conditions when the conditions meet to provide ubiquitous

and seamless V2V and V2I communications in the future intelligent transportation system

(ITS).

103

Chapter 6. Conclusion and Future Works

6.1 Conclusion

In this study, a bi-level MPC-based strategy for connected and automated hybrid electric

vehicles (CAHEVs) was proposed. The main objectives were to design an eco-driving

strategy for CAHEVs that is able to improve fuel economy, minimize engine-out emissions

and realize safe driving under trips with mixed driving scenarios.

In Chapter 2, the foundations of hybrid electric vehicles (HEVs) were introduced, followed

by the description and classification of eco-driving scenarios. Then, the literature on energy

management strategies (EMSs) for HEVs was discussed, emphasizing their principles,

advantages, and limitations. Moreover, the control strategies for connected and automated

hybrid electric vehicles (CAHEVs) were listed and discussed. Finally, the research gap was

pointed out, which paved the way for the introduction to the proposed strategy.

In Chapter 3, a single-shaft, parallel hybrid electric powertrain was developed for the

studied vehicle. As the real-time data sources for decision-making in intelligent

transportation system (ITS), Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I)

communications were then introduced.

In Chapter 4, the problem was described as an eco-driving problem and was then

formulated into a multi-objective optimization problem in the context of the proposed

MPC-based strategy. The objectives are (1) driving safety that aims to comply the vehicle

speed with the road limit, ensure a collision-free following distance and anticipate the

driving behaviour at an intersection over the entire trip, (2) fuel-saving that minimizes the

fuel consumption over the entire trip and (3) exhaust emission reduction that minimizes

the overall engine-out pollutants throughout the trip. The goal of driving safety is achieved

through the use of a Driving Scenario Classifier (DSC) in this intelligent system which

classifies the real-time driving scenario into three categories, namely, free driving scenario,

104

signal anticipation scenario and car-following scenario, such that a corresponding cost

function can be applied in real-time. The optimal control actions are executed on multiple

levels. At the upper level, the real-time vehicle acceleration is determined, while at the

lower level, the optimal torque split ration between ICE and EM is derived. A widely used

rule-based electric-assist control strategy (EACS) for powertrain management is adopted

as the benchmark, and the intelligent driver model (IDM) is used for generating a collision-

free realistic speed profile.

In Chapter 5, the simulation results showed that in a realistic highway-urban trip, the safe

driving goal was achieved throughout the entire driving cycle in the selected urban traffic

environment. Moreover, in comparison to the EACS benchmark strategy, the proposed

MPC-based strategy could also reduce the fuel consumption by 26.80% while keeping the

battery in a healthy range, and reducing the exhaust emissions (𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥) by 27.69%,

73.91% and 47.59%, respectively. Verification was carried out on a suburban-urban trip.

The results from the proposed eco-driving strategy demonstrated even better performance,

as a 38.15% reduction in fuel, 31.08%, 72.72% and 30.73% reduction in the exhaust

emissions (𝐻𝐶, 𝐶𝑂, 𝑁𝑂𝑥), respectively, and the capability to sustain a balanced battery

SOC level throughout the trip, which demonstrated the effectiveness and robustness of the

strategy over a random trip with complex types of driving conditions.

6.2 Future Works

As the strategy is mainly based on the vehicle longitudinal optimization, in the future, it

will also include the dynamic lane-changing strategy to make it more comprehensive. In

addition, the strategy will be improved by introducing the eco-routing algorithm, which

selects the most fuel-efficient route by accurate estimation of the fuel cost on each road in

the dynamic traffic environment. The application of the eco-routing approach in an urban

traffic environment is highly correlated to the prediction accuracy of future driving

conditions such as the traffic density and trends. Moreover, for the sake of realistic

105

application in the real world, the V2V and V2I communication delay will also be

considered in the driving scenarios. Also, the travelling time of the host vehicle can be

taken into consideration in the multi-objective optimization for a more practical driving

experience, and the vehicle driving comfort can be considered to regulate the unnecessary

fluctuation in the speed profile to better serve for human passengers. Besides, the work can

be extended to the optimal control strategy for a CAHEV fleet in the urban driving

conditions, in which case, more data such as the speed and intention of the neighboring

vehicles are required for more efficient and safer coordination between vehicles in the same

fleet. Finally, to make the proposed strategy applicable to the local traffic network in

simulation, 24-hour traffic demands from the local city will be collected and modelled to

build up a realistic urban traffic network for simulation. On the basis of that, to introduce

more realistic conditions into this optimization problem, the locations of gas stations and

electric charging stations can be considered in the urban map.

106

Reference

[1] L. Li, Y. Zhang, C. Yang, X. Jiao, L. Zhang, and J. Song, “Hybrid genetic algorithm-based

optimization of powertrain and control parameters of plug-in hybrid electric bus,” Journal of the Franklin

Institute, vol. 352, no. 3, pp. 776–801, 2015, doi: 10.1016/j.jfranklin.2014.10.016.

[2] W. Enang and C. Bannister, “Modelling and control of hybrid electric vehicles (A comprehensive

review),” Renewable and Sustainable Energy Reviews, vol. 74, pp. 1210–1239, 2017, doi:

10.1016/j.rser.2017.01.075.

[3] C. Guardiola, B. Plá, S. Onori, and G. Rizzoni, “Insight into the HEV/PHEV optimal control

solution based on a new tuning method,” Control Engineering Practice, vol. 29, pp. 247–256, 2014, doi:

10.1016/j.conengprac.2014.01.022.

[4] S. Zhang, T. Niu, Y. Wu, K. M. Zhang, T. J. Wallington, and Q. Xie, “Fine-grained vehicle emission

management using intelligent transportation system data,” Environmental pollution (Barking, Essex :

1987), vol. 241, pp. 1027–1037, 2018, doi: 10.1016/j.envpol.2018.06.016.

[5] U.S. Energy Information Administration (EIA), “ANNUAL ENERGY OUTLOOK 2020,” 2020.

[Online]. Available: https://www.eia.gov/outlooks/aeo/pdf/AEO2020%20Full%20Report.pdf

[6] L. Li, X. Wang, and J. Song, “Fuel consumption optimization for smart hybrid electric vehicle

during a car-following process,” Mechanical Systems and Signal Processing, vol. 87, pp. 17–29, 2017,

doi: 10.1016/j.ymssp.2016.03.002.

[7] D. A. Crolla and D. Cao, “The impact of hybrid and electric powertrains on vehicle dynamics,

control systems and energy regeneration,” Vehicle System Dynamics, vol. 50, sup1, pp. 95–109, 2012,

doi: 10.1080/00423114.2012.676651.

[8] E. Silvaş, T. Hofman, and M. Steinbuch, “Review of Optimal Design Strategies for Hybrid Electric

Vehicles,” IFAC Proceedings Volumes, vol. 45, no. 30, pp. 57–64, 2012, doi: 10.3182/20121023-3-FR-

4025.00054.

[9] M. A. Hannan, F. A. Azidin, and A. Mohamed, “Hybrid electric vehicles and their challenges: A

review,” Renewable and Sustainable Energy Reviews, vol. 29, pp. 135–150, 2014, doi:

10.1016/j.rser.2013.08.097.

[10] M. F. M. Sabri, K. A. Danapalasingam, and M. F. Rahmat, “A review on hybrid electric vehicles

architecture and energy management strategies,” Renewable and Sustainable Energy Reviews, vol. 53,

pp. 1433–1442, 2016, doi: 10.1016/j.rser.2015.09.036.

[11] Y. Huang, H. Wang, A. Khajepour, H. He, and J. Ji, “Model predictive control power management

strategies for HEVs: A review,” Journal of Power Sources, vol. 341, pp. 91–106, 2017, doi:

10.1016/j.jpowsour.2016.11.106.

[12] T. Liu, Y. Zou, D. Liu, and F. Sun, “Reinforcement Learning of Adaptive Energy Management With

Transition Probability for a Hybrid Electric Tracked Vehicle,” IEEE Trans. Ind. Electron., vol. 62, no.

12, pp. 7837–7846, 2015, doi: 10.1109/TIE.2015.2475419.

107

[13] Y. Zou, T. Liu, D. Liu, and F. Sun, “Reinforcement learning-based real-time energy management

for a hybrid tracked vehicle,” Applied Energy, vol. 171, pp. 372–382, 2016, doi:

10.1016/j.apenergy.2016.03.082.

[14] S.-Y. Chen, Y.-H. Hung, C.-H. Wu, and S.-T. Huang, “Optimal energy management of a hybrid

electric powertrain system using improved particle swarm optimization,” Applied Energy, vol. 160, pp.

132–145, 2015, doi: 10.1016/j.apenergy.2015.09.047.

[15] Y. Hu, W. Li, K. Xu, T. Zahid, F. Qin, and C. Li, “Energy Management Strategy for a Hybrid

Electric Vehicle Based on Deep Reinforcement Learning,” Applied Sciences, vol. 8, no. 2, p. 187, 2018,

doi: 10.3390/app8020187.

[16] V. Ngo, T. Hofman, M. Steinbuch, and A. Serrarens, “Predictive gear shift control for a parallel

Hybrid Electric Vehicle,” in 2011 IEEE vehicle power and propulsion conference (VPPC 2011): Chicago,

Illinois, USA 6-9 September 2011, Chicago, IL, USA, 2011, pp. 1–6.

[17] D. Rotenberg, A. Vahidi, and I. Kolmanovsky, “Ultracapacitor Assisted Powertrains: Modeling,

Control, Sizing, and the Impact on Fuel Economy,” IEEE Trans. Contr. Syst. Technol., vol. 19, no. 3,

pp. 576–589, 2011, doi: 10.1109/TCST.2010.2048431.

[18] H. Borhan, A. Vahidi, A. M. Phillips, M. L. Kuang, I. V. Kolmanovsky, and S. Di Cairano, “MPC-

Based Energy Management of a Power-Split Hybrid Electric Vehicle,” IEEE Trans. Contr. Syst.

Technol., vol. 20, no. 3, pp. 593–603, 2012, doi: 10.1109/TCST.2011.2134852.

[19] L. Johannesson, M. Asbogard, and B. Egardt, “Assessing the Potential of Predictive Control for

Hybrid Vehicle Powertrains Using Stochastic Dynamic Programming,” IEEE Trans. Intell. Transport.

Syst., vol. 8, no. 1, pp. 71–83, 2007, doi: 10.1109/TITS.2006.884887.

[20] J. Vales-Alonso, F. Vicente-Carrasco, and J. Alcaraz, “Optimal configuration of roadside beacons

in V2I communications,” Computer Networks, vol. 55, no. 14, pp. 3142–3153, 2011, doi:

10.1016/j.comnet.2011.05.006.

[21] E. Ndashimye, S. Ray, N. Sarkar, and J. Gutiérrez, “Vehicle-to-infrastructure communication over

multi-tier heterogeneous networks: A survey,” Computer Networks, vol. 112, pp. 144–166, 2017, doi:

10.1016/j.comnet.2016.11.008.

[22] D. Jia and D. Ngoduy, “Enhanced cooperative car-following traffic model with the combination of

V2V and V2I communication,” Transportation Research Part B: Methodological, vol. 90, pp. 172–191,

2016, doi: 10.1016/j.trb.2016.03.008.

[23] Y. Wang, D. Zhang, Y. Liu, B. Dai, and L. H. Lee, “Enhancing transportation systems via deep

learning: A survey,” Transportation Research Part C: Emerging Technologies, vol. 99, pp. 144–163,

2019, doi: 10.1016/j.trc.2018.12.004.

[24]K. Jadaan, S. Zeater, and Y. Abukhalil, “Connected Vehicles: An Innovative Transport Technology,”

Procedia Engineering, vol. 187, pp. 641–648, 2017, doi: 10.1016/j.proeng.2017.04.425.

[25] A. Sciarretta and A. Vahidi, “Energy Saving Potentials of CAVs,” in Lecture notes in intelligent

transportation and infrastructure, 2523-3440, Energy-efficient driving of road vehicles: Toward

cooperative, connected, and automated mobility / Antonio Sciarretta, Ardalan Vahidi, A. Sciarretta and

A. Vahidi, Eds., Cham: Springer, 2020, pp. 1–31. [Online]. Available: https://doi.org/10.1007/978-3-

030-24127-8_1

108

[26] J. Wu, X. Wang, L. Li, C.'a. Qin, and Y. Du, “Hierarchical control strategy with battery aging

consideration for hybrid electric vehicle regenerative braking control,” Energy, vol. 145, pp. 301–312,

2018, doi: 10.1016/j.energy.2017.12.138.

[27] L. Guo, B. Gao, Y. Gao, and H. Chen, “Optimal Energy Management for HEVs in Eco-Driving

Applications Using Bi-Level MPC,” IEEE Trans. Intell. Transport. Syst., vol. 18, no. 8, pp. 2153–2162,

2017, doi: 10.1109/TITS.2016.2634019.

[28] W.-Y. Chou, Y.-C. Lin, Y.-H. Lin, and S.-Y. Chen, “Intelligent eco-driving suggestion system based

on vehicle loading model,” in ITS 2012: 2012 12th International Conference on ITS

Telecommunications : 5-8 November 2012 in Tapei, Taiwan, Taipei, Taiwan, 2012, pp. 558–562.

[29] A. Rakotonirainy, G. S. Larue, S. Demmel, and H. Malik, “Fuel consumption and gas emissions of

an automatic transmission vehicle following simple eco-driving instructions on urban roads,” IET

Intelligent Transport Systems, vol. 8, no. 7, pp. 590–597, 2014, doi: 10.1049/iet-its.2013.0076.

[30]I Wengraf, “Easy on the Gas: The effectiveness of eco-driving,” Royal Automobile Club Foundation,

Tech. Rep., 2012.

[31] T.-W. Sung, L.-C. Shiu, F.-T. Lin, and C.-S. Yang, “A Speed Control Scheme of Eco-Driving at

Road Intersections,” in 2015 Third International Conference on Robot, Vision and Signal Processing -

RVSP 2015: 18-20 November 2015, Kaohsiung, Taiwan : proceedings, Kaohsiung, Taiwan, 2015, pp.

51–54.

[32] X. Xiang, K. Zhou, W.-B. Zhang, W. Qin, and Q. Mao, “A Closed-Loop Speed Advisory Model

With Driver's Behavior Adaptability for Eco-Driving,” IEEE Trans. Intell. Transport. Syst., vol. 16, no.

6, pp. 3313–3324, 2015, doi: 10.1109/TITS.2015.2443980.

[33] L. Li, X. Wang, R. Xiong, K. He, and X. Li, “AMT downshifting strategy design of HEV during

regenerative braking process for energy conservation,” Applied Energy, vol. 183, pp. 914–925, 2016,

doi: 10.1016/j.apenergy.2016.09.031.

[34] G. Wu, X. Zhang, and Z. Dong, “Powertrain architectures of electrified vehicles: Review,

classification and comparison,” Journal of the Franklin Institute, vol. 352, no. 2, pp. 425–448, 2015, doi:

10.1016/j.jfranklin.2014.04.018.

[35] K. Çağatay Bayindir, M. A. Gözüküçük, and A. Teke, “A comprehensive overview of hybrid electric

vehicle: Powertrain configurations, powertrain control techniques and electronic control units,” Energy

Conversion and Management, vol. 52, no. 2, pp. 1305–1313, 2011, doi:

10.1016/j.enconman.2010.09.028.

[36] S. J. Moura, D. S. Callaway, H. K. Fathy, and J. L. Stein, “Tradeoffs between battery energy capacity

and stochastic optimal power management in plug-in hybrid electric vehicles,” Journal of Power Sources,

vol. 195, no. 9, pp. 2979–2988, 2010, doi: 10.1016/j.jpowsour.2009.11.026.

[37] S. F. Tie and C. W. Tan, “A review of energy sources and energy management system in electric

vehicles,” Renewable and Sustainable Energy Reviews, vol. 20, pp. 82–102, 2013, doi:

10.1016/j.rser.2012.11.077.

[38] B. Jacobson, “On Vehicle Driving Cycle Simulation,” in SAE Technical Paper Series, 1995.

[39] H. Richter, V. Korte, and L. C. van Beckhoven, “Actual Driving Conditions in Europe and the

Proposed EEC Extra Urban Driving Cycle,” in SAE Technical Paper Series, 1989.

109

[40] L. Berzi, M. Delogu, and M. Pierini, “Development of driving cycles for electric vehicles in the

context of the city of Florence,” Transportation Research Part D: Transport and Environment, vol. 47,

pp. 299–322, 2016, doi: 10.1016/j.trd.2016.05.010.

[41] B. Liu, Q. Shi, L. He, and D. Qiu, “A study on the construction of Hefei urban driving cycle for

passenger vehicle,” IFAC-PapersOnLine, vol. 51, no. 31, pp. 854–858, 2018, doi:

10.1016/j.ifacol.2018.10.100.

[42] C. Kim, E. NamGoong, S. Lee, T. Kim, and H. Kim, “Fuel Economy Optimization for Parallel

Hybrid Vehicles with CVT,” in SAE Technical Paper Series, 1999.

[43] J. Yang and G. G. Zhu, “Predictive Boundary Management Control of a Hybrid Powertrain,” in

Proceedings of the ASME 7th annual dynamic systems and control conference 2014: October 22-24,

2014, San Antonio, Texas, USA, San Antonio, Texas, USA, 2014.

[44] N. Jalil, N. A. Kheir, and M. Salman, “A rule-based energy management strategy for a series hybrid

vehicle,” in Proceedings of the 1997 American control conference, Albuquerque, NM, USA, 1997, 689-

693 vol.1.

[45] V. H. Johnson, K. B. Wipke, and D. J. Rausen, “HEV Control Strategy for Real-Time Optimization

of Fuel Economy and Emissions,” SAE Technical Paper 2000-01-1543, Apr. 2000. [Online]. Available:

https://www.sae.org/gsdownload/?prodCd=2000-01-1543

[46] F. R. Salmasi, “Control Strategies for Hybrid Electric Vehicles: Evolution, Classification,

Comparison, and Future Trends,” IEEE Trans. Veh. Technol., vol. 56, no. 5, pp. 2393–2404, 2007, doi:

10.1109/TVT.2007.899933.

[47] H.-D. Lee and S.-K. Sul, “Fuzzy-logic-based torque control strategy for parallel-type hybrid electric

vehicle,” IEEE Trans. Ind. Electron., vol. 45, no. 4, pp. 625–632, 1998, doi: 10.1109/41.704891.

[48] A. Rajagopalan, G. Washington, G. Rizzoni, and Y. Guezennec, “Development of Fuzzy Logic and

Neural Network Control and Advanced Emissions Modeling for Parallel Hybrid Vehicles,” 2003.

[49] S. Ichikawa et al., “Novel energy management system for hybrid electric vehicles utilizing car

navigation over a commuting route,” in 2004 IEEE intelligent vehicles symposium, Parma, Italy, 2004,

pp. 161–166.

[50] C.-C. Lin, H. Peng, J. W. Grizzle, and J.-M. Kang, “Power management strategy for a parallel hybrid

electric truck,” IEEE Trans. Contr. Syst. Technol., vol. 11, no. 6, pp. 839–849, 2003, doi:

10.1109/TCST.2003.815606.

[51] C.-C. Lin, H. Peng, and J. W. Grizzle, “A stochastic control strategy for hybrid electric vehicles,”

in 2004 American control conference, Boston, MA, USA, 2004, 4710-4715 vol.5.

[52] B. Huang, Z. Wang, and Y. Xu, “Multi-Objective Genetic Algorithm for Hybrid Electric Vehicle

Parameter Optimization,” in IEEE/RSJ International Conference on Intelligent Robots and Systems,

2006: Oct. 2006, [Beijing, China], Beijing, China, 2006, pp. 5177–5182.

[53] M. Huang and H. Yu, “Optimal Multilevel Hierarchical Control Strategy for Parallel Hybrid Electric

Vehicle,” in IEEE Vehicle Power and Propulsion Conference, 2006: VPPC '06 ; 6-8 Sept. 2006,

[Windsor, United Kingdom], Windsor, UK, 2006, pp. 1–4.

110

[54] Z. Chen, C. C. Mi, B. Xia, and C. You, “Energy management of power-split plug-in hybrid electric

vehicles based on simulated annealing and Pontryagin's minimum principle,” Journal of Power Sources,

vol. 272, pp. 160–168, 2014, doi: 10.1016/j.jpowsour.2014.08.057.

[55] Namwook Kim, Daeheung Lee, Suk-Won Cha,Huei Peng, (PDF) Optimal Control of a Plug-In

Hybrid Electric Vehicle (PHEV) Based on Driving Patterns. [Online]. Available: https://

www.researchgate.net/publication/267718169_Optimal_Control_of_a_Plug-In_Hybrid_Electric_

Vehicle_PHEV_Based_on_Driving_Patterns (accessed: Feb. 23 2020).

[56] G. Paganelli, T. M. Guerra, S. Delprat, J.-J. Santin, M. Delhom, and E. Combes, “Simulation and

assessment of power control strategies for a parallel hybrid car,” Proceedings of the Institution of

Mechanical Engineers, Part D: Journal of Automobile Engineering, vol. 214, no. 7, pp. 705–717, 2000,

doi: 10.1243/0954407001527583.

[57] A. Sciarretta, M. Back, and L. Guzzella, “Optimal Control of Parallel Hybrid Electric Vehicles,”

IEEE Trans. Contr. Syst. Technol., vol. 12, no. 3, pp. 352–363, 2004, doi: 10.1109/TCST.2004.824312.

[58] K. R. Bouwman, T. H. Pham, S. Wilkins, and T. Hofman, “Predictive Energy Management Strategy

Including Traffic Flow Data for Hybrid Electric Vehicles,” IFAC-PapersOnLine, vol. 50, no. 1, pp.

10046–10051, 2017, doi: 10.1016/j.ifacol.2017.08.1775.

[59] C. Sun, S. J. Moura, X. Hu, J. K. Hedrick, and F. Sun, “Dynamic Traffic Feedback Data Enabled

Energy Management in Plug-in Hybrid Electric Vehicles,” IEEE Trans. Contr. Syst. Technol., vol. 23,

no. 3, pp. 1075–1086, 2015, doi: 10.1109/TCST.2014.2361294.

[60] Z. Xiuzheng, Z. Liguo, and Y. Kholodov, “Model predictive control of eco-driving for transit using

V2I communication,” in Control Conference (CCC), 2015 34th Chinese, Hangzhou, China, Jul. 2015 -

Jul. 2015, pp. 2511–2516.

[61] H. Bouvier, G. Colin, and Y. Chamaillard, “Determination and comparison of optimal eco-driving

cycles for hybrid electric vehicles,” in Control Conference (ECC), 2015 European, Linz, Austria, Jul.

2015 - Jul. 2015, pp. 142–147.

[62] V. T. Minh and A. A. Rashid, “Modeling and model predictive control for hybrid electric vehicles,”

Int.J Automot. Technol., vol. 13, no. 3, pp. 477–485, 2012, doi: 10.1007/s12239-012-0045-0.

[63] K. Yu, J. Yang, and D. Yamaguchi, “Model predictive control for hybrid vehicle ecological driving

using traffic signal and road slope information,” Control Theory Technol., vol. 13, no. 1, pp. 17–28,

2015, doi: 10.1007/s11768-015-4058-x.

[64] S.-Y. Shieh, T. Ersal, and H. Peng, “Pulse-and-Glide Operation for Parallel Hybrid Electric Vehicles

with Step-Gear Transmission in Automated Car-Following Scenario with Ride Comfort Consideration,”

in 2019 American Control Conference (ACC), Philadelphia, PA, USA, 2019, pp. 959–964.

[65] S. Zhao, Y. Wu, Y. Zhai, C. Chen, J. Wu, and B. Sun, “An integrated control strategy for eco-driving

of intelligent hybrid vehicles,” in 2019 IEEE International Conference on Smart Manufacturing,

Industrial & Logistics Engineering (SMILE), Hangzhou, China, Apr. 2019 - Apr. 2019, pp. 14–18.

[66] B. Zhang, H. Lim, S. Xu, and W. Su, “Distance-oriented hierarchical control and ecological driving

strategy for HEVs,” IET Electrical Systems in Transportation, vol. 9, no. 1, pp. 44–52, 2019, doi:

10.1049/iet-est.2018.5044.

111

[67] A. Sciarretta and A. Vahidi, Eds., Energy-efficient driving of road vehicles: Toward cooperative,

connected, and automated mobility / Antonio Sciarretta, Ardalan Vahidi. Cham: Springer, 2020.

[68] X. Hu, X. Zhang, X. Tang, and X. Lin, “Model predictive control of hybrid electric vehicles for fuel

economy, emission reductions, and inter-vehicle safety in car-following scenarios,” Energy, vol. 196, p.

117101, 2020, doi: 10.1016/j.energy.2020.117101.

[69] L. Liu, F. Kong, X. Liu, Y. Peng, and Q. Wang, “A review on electric vehicles interacting with

renewable energy in smart grid,” Renewable and Sustainable Energy Reviews, vol. 51, pp. 648–661,

2015, doi: 10.1016/j.rser.2015.06.036.

[70] C. Yi, B. Epureanu, S. K. Hong, T. Ge, and X. G. Yang, “Modeling, control, and performance of a

novel architecture of hybrid electric powertrain system,” Applied Energy, vol. 178, pp. 454–467, 2016,

doi: 10.1016/j.apenergy.2016.06.068.

[71] B. Liu, L. Li, X. Wang, and S. Cheng, “Hybrid Electric Vehicle Downshifting Strategy Based on

Stochastic Dynamic Programming During Regenerative Braking Process,” IEEE Trans. Veh. Technol.,

vol. 67, no. 6, pp. 4716–4727, 2018, doi: 10.1109/TVT.2018.2815518.

[72] J. Y. Wong, Theory of ground vehicles, 4th ed. Hoboken, N.J.: Wiley; Chichester : John Wiley

[distributor], 2008.

[73] J. Lee and D. J. Nelson, “Rotating Inertia Impact on Propulsion and Regenerative Braking for

Electric Motor Driven Vehicles,” in 2005 IEEE Vehicle Power and Propulsion Conference, Chicago, IL,

USA, Sep. 2005, pp. 308–314.

[74] M. Ehsani, Y. Gao, S. Longo, and K. Ebrahimi, Modern electric, hybrid electric, and fuel cell

vehicles, 3rd ed. Boca Raton: CRC Press/Taylor & Francis Group, 2019.

[75] M. Treiber and A. Kesting, Traffic flow dynamics: Data, models and simulation. Heidelberg, New

York: Springer, 2013. [Online]. Available: https://www.eia.gov/outlooks/aeo/

[76] C. Osornio-Correa, R. C. Villarreal-Calva, J. Estavillo-Galsworthy, A. Molina-Cristóbal, and S. D.

Santillán-Gutiérrez, “Optimization of Power Train and Control Strategy of a Hybrid Electric Vehicle for

Maximum Energy Economy,” Ingeniería, Investigación y Tecnología, vol. 14, no. 1, pp. 65–80, 2013,

doi: 10.1016/S1405-7743(13)72226-1.

[77] H. Noori, “Modeling the impact of VANET-enabled traffic lights control on the response time of

emergency vehicles in realistic large-scale urban area,” in IEEE International Conference on

Communications workshops (ICC), 2013: 9-13 June 2013, Budapest, Hungary, Budapest, Hungary,

2013, pp. 526–531.

[78] K. B. Wipke, M. R. Cuddy, and S. D. Burch, “ADVISOR 2.1: a user-friendly advanced powertrain

simulation using a combined backward/forward approach,” IEEE Trans. Veh. Technol., vol. 48, no. 6,

pp. 1751–1761, 1999, doi: 10.1109/25.806767.

[79] P. A. Lopez et al., “Microscopic Traffic Simulation using SUMO,” in 2018 21st International

Conference on Intelligent Transportation Systems (ITSC), Maui, HI, Nov. 2018 - Nov. 2018, pp. 2575–

2582.

[80] S. Uppoor and M. Fiore, “Characterizing Pervasive Vehicular Access to the Cellular RAN

Infrastructure: An Urban Case Study,” IEEE Trans. Veh. Technol., vol. 64, no. 6, pp. 2603–2614, 2015,

doi: 10.1109/TVT.2014.2343651.

112

[81] J. Erdmann, “1 Lane-Changing Model in SUMO,” 2014. [Online]. Available: https://

pdfs.semanticscholar.org/5f36/bc7be59fd4546a5f3b5de4cd91f43bf11825.pdf

[82] J. Song, Y. Wu, Z. Xu, and X. Lin, “Research on car-following model based on SUMO,” in

Proceedings of 2014 IEEE 7th International Conference on Advanced Infocomm Technology

(IEEE/ICAIT2014): November 14-16, 2014, Fuzhou, China, Fuzhou, China, IEEE/ICAIT2014, pp. 47–

55.

development of eco-driving control system for connected

Documents